Fertility, Living Arrengements, Care and Mobilitie - Hohn Stillwell, Ernestina Coast, Dylan Kneale

Fertility, Living Arrangements, Care and Mobility
Understanding Population Trends and Processes
Volume 1
Series Editor
J. Stillwell
In western Europe and other developed parts of the world, there are some very
significant demographic processes taking place at the individual, household, com-
munity and national scales including the ageing of the population, the delay in
childbearing, the rise in childlessness, the increase in divorce, the fall in marriage
rates, the increase in cohabitation, the increase in mixed marriages, the change in
household structures, the rise in step-parenting, and the appearance of new streams
of migration taking place both within and between countries. The relationships
between demographic change, international migration, labour and housing market
dynamics, care provision and intergenerational attitudes are complex to understand
and yet it is vital to quantify the trends and to understand the processes. Similarly, it
is critical to appreciate what the policy consequences are for the trends and processes
that have become apparent. This series has its roots in understanding and analyzing
these trends and processes.
This series will be of interest to a wide range of individuals with interests in demo-
graphic and social change, including demographers, population geographers, soci-
ologists, economists, political scientists, epidemiologists and health researchers and
well as practitioners and commentators across the social sciences.
For further volumes:

http://www.springer.com/series/8113
Fertility, Living Arrangements,
Care and Mobility
Understanding Population Trends

and Processes - Volume 1
Edited by
John Stillwell
University of Leeds, UK
Ernestina Coast
London School of Economics, UK
and
Dylan Kneale
Institute of Education, University of London, UK
123
Editors
Professor John Stillwell Dr. Ernestina Coast
University of Leeds London School of Economics
School of Geography Dept. Social Policy
Fac. Earth and Environment Hougton St.
Woodhouse Lane London
Leeds United Kingdom WC2A 2AE
United Kingdom LS2 9JT e.coast@lse.ac.uk
j.c.h.stillwell@leeds.ac.uk
Dylan Kneale
University of London
Inst. Education
20 Bedford Way
London
United Kingdom WC1H 0AL
D.Kneale@ioe.ac.uk
ISSN 8113
ISBN 978-1-4020-9681-5 e-ISBN 978-1-4020-9682-2
DOI 10.1007/978-1-4020-9682-2
Springer Dordrecht Heidelberg London New York
Library of Congress Control Number: 2008942396
c Springer Science + Business Media B.V. 2009

No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by
any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written
permission from the Publisher, with the exception of any material supplied specifically for the purpose
of being entered and executed on a computer system, for exclusive use by the purchaser of the work.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Foreword
One of the great undertapped sources of social science research has been the
plethora of data from wonderful surveys and censuses, available for secondary
analysis through the Economic and Social Data Service or elsewhere, which do
not receive the rich and thorough analysis they deserve.
The Economic and Social Research Council (ESRC) has attempted to address
this in two ways: through a sustained programme of methods development led by
the National Centre for Research Methods and by developing the capacity of social
scientists to undertake the secondary analysis of large data sets. This book contains a
series of essays from some of the researchers on the very successful ‘Understanding
Population Trends and Processes’ (UPTAP) scheme.
These essays are on linked themes of fertility, living arrangements, care and
mobility which are central to one of ESRC’s key priorities. As UK society becomes
ever more diverse, understanding the seismic changes occurring in the UK pop-
ulation has become increasingly more important and the need to develop policy
responses has become more urgent. This excellent book provides some exciting and
thoughtful insights into some important areas of both research and policy formation.
Finally I would like to thank John Stillwell, who has directed the programme with
alacrity and drive for all his work, not only in the programme but, with Ernestina
Coast and Dylan Kneale, on the production of this excellent volume, the first of a
series on Understanding Population Trends and Processes.
Swindon Ian Diamond

September 2008
v
Preface
The world is changing rapidly and population dynamics are a hugely important
dimension of global transformation. Demographic restructuring is one of the big
challenges of the twenty-first century. Not only do the components and complexion
of population change provide a fascinating arena for the research community, they
present questions of critical importance for practitioners and policy makers.
This book is the first in a series on ‘Understanding Population Trends and Pro-
cesses’, all of which are based on research contributions to our knowledge of differ-
ent aspects of population structure and distribution, all of which involve the analysis
of secondary data from censuses, surveys or administrative records, and all of which
report results for Britain (and elsewhere in Europe in some cases).
This volume brings together a series of studies that focus particularly on the
household and the roles of adults and children. Particular attention is paid to studies
of bearing and raising children, of living together or alone, of caring for those
in need, and of moving home or changing school – hence the four cross-cutting
themes that constitute the subtitle of the book – fertility, living arrangements, care
and mobility.
Leeds, UK John Stillwell

London, UK Ernestina Coast
London, UK Dylan Kneale
September 2008
vii
Acknowledgements
All the research reported in this volume has been funded by the UK Economic and
Social Research Council (ESRC) as part of the ‘Understanding Population Trends
and Processes’ (UPTAP) programme. John Stillwell, the programme director, is
grateful for the support he has received from staff at the ESRC, particularly Jeremy
Neathy, Mike Bright, Jo Duffy, Jennifer Edwards and Heath Bampton.
The editors would like to thank the contributors for providing draft chapters
on time and responding to editorial comments so promptly. Thanks also to David
Appleyard in the Graphics Unit at the School of Geography, University of Leeds
who has done a great job in redrawing all the figures in a consistent style.
ix
Contents
1 Fertility, Living Arrangements, Care and Mobility . . . . . . . . . . . . . . . . . 1

Dylan Kneale, Ernestina Coast and John Stillwell
2 Delayed Childbearing and Childlessness . . . . . . . . . . . . . . . . . . . . . . . . . . 23

Roona Simpson
3 Women’s Education and Childbearing: A Growing Divide . . . . . . . . . . 41

Sarah Smith and Anita Ratcliffe
4 The Timing of Motherhood, Mothers’ Employment and Child

Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Kirstine Hansen, Denise Hawkes and Heather Joshi
5 Early Parenthood: Definition and Prediction in Two British Cohorts . 81

Dylan Kneale
6 Currently Cohabiting: Relationship Attitudes, Expectations and

Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Ernestina Coast
7 Living Arrangements, Health and Well-Being . . . . . . . . . . . . . . . . . . . . . . 127

Harriet Young and Emily Grundy
8 Stepparenting and Mental Health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

Peteke Feijten, Paul Boyle, Zhiqiang Feng, Vernon Gayle
and Elspeth Graham
9 Grandparents and the Care of Their Grandchildren . . . . . . . . . . . . . . . . 171

Alison Smith Koslowski
xi
xii Contents
10 Internal Migration and Inter-household Relationships . . . . . . . . . . . . . . 191

Oliver Duke-Williams
11 Exploring Dimensions of School Change During Primary Education

in England . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Joan Wilson
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Contributors
Paul Boyle School of Geography and Geosciences, University of St Andrews, St

Andrews KY16 9AL, UK; Longitudinal Studies Centre – Scotland, University of
St Andrews, UK, p.boyle@st-andrews.ac.uk
Ernestina Coast Department of Social Policy, London School of Economics,

Houghton Street, London, WC2A 2AE, UK, E.Coast@lse.ac.uk
Oliver Duke-Williams School of Geography, University of Leeds, Leeds LS2

9JT, UK, o.w.duke-williams@leeds.ac.uk
Peteke Feijten School of Geography and Geosciences, University of St Andrews,

St Andrews KY16 9AL, UK; Longitudinal Studies Centre – Scotland, University
of St Andrews, UK, pmf1@st-andrews.ac.uk
Zhiqiang Feng School of Geography and Geosciences, University of St Andrews,

St Andrews KY16 9AL, UK; Longitudinal Studies Centre – Scotland, University
of St Andrews, UK, zf2@st-andrews.ac.uk
Vernon Gayle Department of Applied Social Science, University of Stirling,

Stirling FK9 4LA, Scotland, UK, Vernon.gayle@stir.ac.uk
Elspeth Graham School of Geography and Geosciences, University of St

Andrews, St Andrews KY16 9AL, UK, efg@st-andrews.ac.uk
Emily Grundy Centre for Population Studies, London School of Hygiene and
Tropical Medicine, University of London, UK, Emily.Grundy@lshtm.ac.uk
Kirstine Hansen Centre for Longitudinal Studies, Institute of Education,

University of London, London, UK, K.Hansen@ioe.ac.uk
Denise Hawkes Business School, University of Greenwich, London, UK and

Centre for Longitudinal Studies, Institute of Education, University of London,
London, UK, D.Hawkes@gre.ac.uk
Heather Joshi Centre for Longitudinal Studies, Institute of Education, University

of London, London, UK, H.Joshi@ioe.ac.uk
xiii
xiv Contributors
Dylan Kneale Centre for Longitudinal Studies, Institute of Education, University

of London, London, UK, D.Kneale@ioe.ac.uk
Anita Ratcliffe Department of Economics and Centre for Market and Public
Organisation, University of Bristol, Bristol, UK, anita.ratclife@bristol.ac.uk
Roona Simpson Centre for Research on Families and Relationships, University of
Edinburgh, UK, Roona.Simpson@ed.ac.uk
Sarah Smith Department of Economics and Centre for Market and Public
Organisation, University of Bristol, Bristol, UK, sarah.smith@bristol.ac.uk
Alison Smith Koslowski School of Social and Political Science, University of
Edinburgh, Edinburgh EH8 9LD, UK, Alison.J.Smith@ed.ac.uk
John Stillwell School of Geography, University of Leeds, Leeds LS2 9JT, UK,
j.c.h.stillwell@leeds.ac.uk
Joan Wilson Centre for Longitudinal Studies, Institute of Education, University
of London; Centre for the Economics of Education; and Centre for Economic
Performance, London School of Economics, London, UK, JWilson@ioe.ac.uk
Harriet Young Centre for Population Studies, London School of Hygiene and
Tropical Medicine University of London, UK, Harriet.Young@lshtm.ac.uk
Chapter 1
Fertility, Living Arrangements, Care
and Mobility
Dylan Kneale, Ernestina Coast and John Stillwell
Introduction
Demographic change constitutes one of the most important challenges of the
twenty-first century. Population ageing has become the focus of attention for ana-
lysts seeking to establish its causes and consequences and policy makers charged
with responding to its implications. The proportion of older people is expanding
dramatically due to declining fertility and improving mortality.
In this book we are concerned with understanding the trends and processes occur-
ring, particularly but not exclusively, in the earlier rather than the later stages of
the life course. By assembling a number of research studies on the processes sur-
rounding fertility and the patterns of living arrangements that characterise society
in the new millennium, we hope to provide novel and detailed insights into socio-
demographic change in the United Kingdom. However, the contents of the book
also reflect two other key themes that have become increasingly important in recent
years and are set to become even more so in the future. The first of these themes is
that of care – not only that of the elderly by family members or other providers, but
also care of children, sometimes by elderly relatives, when parents return to work,
for example. According to United Nations projections (United Nations, 2005), the
ratio of people of working age to those of non-working age, currently just over two
for Europe as a whole, is projected to fall dramatically over the coming decades with
severe implications for both the demand for care and care provision. The final theme
is population mobility, a concept that embraces a series of behaviours at different
spatial scales including international migration, residential mobility and daily com-
muting. International migration is clearly of fundamental significance on a global
scale with major pressures mounting on Europe from the developing world (see
Holzmann and Münz, 2004, for example). In this volume, the focus is much more
localised, concentrating on the movement of different household types in the UK
D. Kneale (B)
Centre for Longitudinal Studies, Institute of Education, University of London, London, UK
e-mail: D.Kneale@ioe.ac.uk
J. Stillwell et al. (eds.), Fertility, Living Arrangements, Care and Mobility, Understanding 1
Population Trends and Processes 1, DOI 10.1007/978-1-4020-9682-2 1,

C Springer Science+Business Media B.V. 2009
2 D. Kneale et al.
together with mobility of children between schools, topics which complement the
chapters of the book on child-bearing and living arrangements.
This introductory chapter establishes a context for the chapters which follow,
discussing each of the major themes on which the book is based and underscoring
some of the key conclusions of the projects reported by the contributing authors,
all of which have been undertaken under the umbrella of the Economic and Social
Research Council’s programme on Understanding Population Trends and Processes
(UPTAP). All the chapters of the book contain the results of analysis of secondary
data, the methodological requirement of UPTAP, and therefore we end the chapter
with a summary of the data sets that have been utilised by contributing authors. We
begin, in the next section, with fertility.
Fertility
Striking patterns of low and late fertility are now firmly entrenched within British
demography, and are replicated across most developed countries. While movement
towards, and maintenance of, relatively low fertility has been present since the turn
of the twentieth century as a result of the first demographic transition (Jefferies,
2005), it is the change in the determinants of fertility that has been most noteworthy.
Following the first demographic transition in which death rates and birth rates
moved over time from high to low levels, the second demographic transition theory
outlines the changes said to be representative of the continuation of falling fertil-
ity rates – below replacement level fertility, growing old-age dependency ratios and
decreasing child-dependency ratios – as well as outlining the determinants of these
changes (Van de Kaa, 1987; Lesthaeghe and Neels, 2002). In particular, the notion
of the second demographic transition describes a complement of social changes that
lead to lower fertility including declining marriage rates, increasing divorce and
cohabitation, pluralistic household structures and increased female participation in
both higher education and the labour force (Van de Kaa, 1987). One of the salient
characteristics of this transition is increasing age at first parenthood (Lesthaeghe
and Neels, 2002) which is both predicted by the determinants of the transition, and
in itself is a marker of lower fertility. Much attention in the chapters examin-
ing fertility in this book is focused upon age at first parenthood, both in terms
of focusing on the predictors of age at family formation, as well as on the
effects that parental age and other indicators of the second demographic transition,
such as changing family structures and increasing mother’s employment, have on
children.
Statistics to evidence lower and later fertility in the UK are in no short supply.
Since the 1970s, fertility has remained below replacement level (Smallwood and
Chamberlain, 2005) and the total fertility rate has deviated little from around 1.7
children per woman in the past two decades up to 2006. While the level of fertility
has largely remained constant over this time, the age at motherhood has continued to
rise since the mid 1970s, from an average of 26.5 years in 1976 to almost 29 years
1 Fertility, Living Arrangements, Care and Mobility 3
Fig. 1.1 Age-specific 60 29.5
Births per 1,000 women

fertility rates (ASFR) for 29
Average age (Years)

50
women aged 15–19 and 28.5
35–39 and average age at 40 28
motherhood 27.5
(Source: Office for National 30
27
Statistics, 2007)
20 26.5
ASFR 15–19 26
10 ASFR 53–39
Average age motherhood 25.5
0 25
1961
1965
1969
1973
1977
1981
1985
1989
1993
1997
2001
2005
Years
in 2006 (Office for National Statistics, 2007). Such a rise has not been observed
uniformly across fertility schedules and represents some divergence for different
groups of women.
Most of this rise in age at first motherhood can be seen to originate from the
decreasing rates of entry into motherhood in the twenties and rises in older age fer-
tility. There is an overall ‘flight from parenthood’ in the twenties, with many women
postponing motherhood until their mid thirties. However, while later motherhood
has become the trend for the majority of women, early and teenage motherhood
persists for the minority (Hadfield et al., 2007). Figure 1.1 shows the maintenance
of teenage age-specific fertility rates since the late 1970s despite several recent inter-
ventions (Social Exclusion Unit, 1999), with the UK identified as having the highest
rates of teenage motherhood in western Europe (UNICEF, 2001). The average age
at motherhood has also been increasing consistently since the late 1970s on a sim-
ilar trajectory to that of the age-specific birth rate for women aged 35–39 years.
Recently, the effects of postponement and rising age at first birth have been cause
for concern for policy makers; although it is early motherhood that has remained as
a prime concern where fertility is in question.
In this case, it is not the young age of the mother per se that is of concern to
policy makers and academics, but the differences in the characteristics of women
having children early and the outcomes of their children, particularly when com-
pared to older women. This has been described as a process of social polarisation
in entry to parenthood (Joshi, 2007) and is addressed in this volume by describ-
ing the characteristics, extent and effects of social polarisation in the time to first
birth. Education has been identified as a key driver of postponement. Highly edu-
cated women are found to be those delaying family formation the most, having
most to lose from time out of the labour market (Berrington, 2004; Gonzalez and
Jurado-Guerrero, 2006; Rendall et al., 2005; Rendall and Smallwood, 2003); and
educational class is perhaps the most important marker of this social polarisa-
tion. Polarisation in age of first motherhood and the relationship with education
4 D. Kneale et al.
represent an overarching theme in all the chapters that examine fertility in this
volume.
Postponement and childlessness are examined in Roona Simpson’s analysis
of differences across two British birth cohorts reported in Chapter 2. Here, the
characteristics of those remaining childless are compared with those who have
entered parenthood. Simpson’s research indicates that while both men and women
are increasingly delaying transition to parenthood, men in particular are postponing
transition to fatherhood. She also confirms the findings of other studies that show
that women from lower social class backgrounds and those who hold lower educa-
tional qualifications are also those entering parenthood first. This work represents
one of a growing number of works that seek to redress the gender imbalance in
the majority of studies examining the determinants and markers of fertility, through
examining patterns for both men and women. However, she also finds the same
result among men and her chapter describes some of the polarised patterns of entry
into fatherhood.
As discussed previously, postponement (and childlessness) is usually associated
with high levels of education and strong ties to the labour market. A key issue that
is addressed in this volume is how this attachment to the labour market has changed
over time and its impact on entering motherhood. The analyses presented not only
look at educational level and labour market participation as a predictor of entry
into motherhood, but also at patterns after motherhood. Labour market participation
is increasingly compatible with motherhood (Joshi, 2002; Edwards, 2002) and in
fact, a growing number of mothers do find themselves working, either by choice or
through necessity. However, working mothers have traditionally returned to work
on a part-time basis, and have been concentrated in low paid, gender segregated
work (Joshi, 2002; Dex et al., 1998; Coyle, 2005). Recent years have seen the intro-
duction of family friendly workplace environments and policies, such as maternity
leave allowance (Dex et al., 1998; Lewis and Campbell, 2007) and paternity leave,
which may facilitate balancing work and motherhood. The analyses presented in
this volume present a detailed description of working practices of mothers and in
some ways, are a reflection of the impact of such policies. The implication of such
increases in the numbers of working mothers is that childcare moves out of the sole
domain of the mother; and for those who re/enter the labour market a range of child-
care options growing in availability and diversity are available (Lewis and Campbell,
2007). Despite this growth, not all options may be available to every woman; it is
highly educated women with strong ties to the labour market who are also those
able to negotiate childcare arrangements with partners and relatives and who are
able to purchase childcare elsewhere (Coyle, 2005). This, again suggests a slightly
cyclical pattern whereby highly educated women have initially stronger ties to the
workplace leading to the postponement of births; but are also those correspondingly
who are able to reengage with the labour market, more often than not on a full-time
basis.
In Chapter 3 on the effect of women’s education on time to first motherhood,
Sarah Smith and Anita Ratcliffe address the issue of polarised entry into mother-
hood but also consider in more detail the effect of this polarisation on mothers’
employment and childcare practices. They examine the literature on the links
between education and entry into motherhood, showing a negative correlation gen-
erally, but they also demonstrate that the welfare state can buffer this association,
and narrow the discrepancy between births among highly educated and less edu-
cated women. Differentials by educational level also extend into child care, where
highly qualified women are also those most likely to rely upon formal childcare.
Their results distinguish between those who left education at the minimum age,
those who left at age 18 and those who left at 18 and went on to higher educa-
tion, indicating that experience of any higher education appears to be particularly
associated with postponement and childlessness.
In Chapter 4, Kirstine Hansen, Denise Hawkes and Heather Joshi begin by
examining the age at first motherhood, finding that this increased between all three
cohorts they study and that educational level remained the strongest predictor of
transition to first parenthood. However, they also note the importance of child-
hood disadvantage as a predictor, with disadvantage propelling younger women
into motherhood. Having established the link between education and timing of
motherhood, and how educational level will influence labour market participation
and advantage, Hansen et al. move to examine labour market participation among
mothers specifically. As with Smith and Ratcliffe, they find an increased propen-
sity among more recent cohorts of mothers to be employed. They also similarly
find that higher qualified mothers are more likely to enter employment. However,
they also find previous engagement in the labour market to be a strong predictor.
Finally, after illuminating the links between the age of the mother, her education,
and her labour market participation, they move on to examine childcare. They intro-
duce analyses that highlight differences between modes of childcare, educational
class and labour market participation. In addition the analysis goes one step fur-
ther by assessing the quality of this childcare in terms of child outcomes, giving an
indication of the implications of both maternal employment and maternal choices in
childcare.
Polarised pathways to motherhood are explored further in this volume by examin-
ing the other end of the fertility spectrum – early parenthood. Teenage motherhood
has long been associated with a range of negative characteristics including poor
educational background, poverty and welfare dependence, and an unstable fam-
ily life (summarised, for example, in Imamura et al., 2007; Harden et al., 2006).
While teenage mothers are associated with these disadvantaged characteristics; a
body of evidence suggests that there are negligible benefits for these mothers in
delaying parenthood (Hotz et al., 2004; Goodman et al., 2004). This is mainly due
to teenage motherhood being a marker, as opposed to a cause, of disadvantage;
and concentrating childbearing at earlier points may actually be a beneficial strat-
egy in terms of labour market opportunities for this specific group of women. But
despite the questionable evidence as to teenage parenthood’s status as a cause, as
opposed to a marker, of poverty, this group has been the focus of several policy
interventions. In Chapter 5, Dylan Kneale offers a short discussion on the politici-
sation of the term ‘teenage’ parents and questions why the under 20 and over 20
years cut-off has remained such a pervasive term for a diminishing group of parents.
6 D. Kneale et al.
In acknowledging that some of the focus may be justified in terms of outcomes for
children, Kneale sets about examining the pre-existing characteristics of early par-
ents to examine continuities between early parents into their early twenties and those
aged under twenty. Continuities in terms of known and hypothesised predictors of
the timing to young parenthood found in the literature are examined. Kneale high-
lights the strength of tenure over social class as a predictor of both early motherhood
and fatherhood. Results on the effect of dislike of school and family building pref-
erences as predictors of early parenthood are also presented. The chapter concludes
that while there is not a structural break in the characteristics of teenage mothers
compared to mothers in their early twenties, there is sufficient evidence to speculate
that teenage fathers do actually represent a distinct group away from other fathers
in their early twenties.
A key issue addressed in all of the chapters on fertility, and significant for policy
makers and academics alike, is whether the socially polarised divide in reproductive
timing is growing. Each of the chapters on fertility is able to inform on this issue
specifically by analysing the reproductive behaviour of different cohorts of women,
as opposed to taking a period approach. Most of the research takes a longitudi-
nal, lifecourse approach through either examining childhood factors as predictors
of later life fertility or through examining detailed occupational, educational and
partnership histories of women. This gives much of the research on fertility con-
tained within the volume a degree of insight that is absent from many other studies,
through including information that would otherwise be impossible to obtain because
of bias or recall error. This approach also allows for links to be made between indi-
viduals under study and the historical context and social structure present (Elliott,
2005). Individuals included in these analyses would have been subject to some major
changes in terms of education with the raising of the school age participation in
higher education (Power and Elliott, 2006), increasing equality in the workplace
through legislation such as the Equal Pay Act (Dex et al., 1998), a move towards
family friendly policies such as maternity leave (Dex et al., 1998), but also poli-
cies proscribing right and wrong pathways to motherhood (for example, Social
Exclusion Unit, 1999).
Finally in this section, it is important to recognise that men have often been
neglected in studies of fertility. In most cases, this has been because of either a
lack of data or because of questionable reliability of male accounts of fertility
(Rendall et al., 1999; Darroch et al., 1999; Greene and Biddlecom, 2000). This com-
parative lack of research, both on an intuitive and evidential level, does not reflect
the importance of fatherhood (see for example, Pleck, 2007, Sarkadi et al., 2008). It
is hoped that the results presented in this volume can make a contribution to fam-
ily building policies and knowledge, said to have suffered thus far from the lack
of input of male fertility histories (Flood, 2007). The results from Chapters 2–5
will also make an important contribution to knowledge on recent polarisations in
reproduction and the multifaceted effects on age at first birth, labour market partic-
ipation, childcare patterns and child outcomes. In the next section, issues relating
specifically to changing household and family structures are discussed as we move
from discussing who has children and when, to living arrangements in which chil-
dren also play a significant role.
Living Arrangements
The ways in which people organise their living arrangements are both causes and
consequences of social and societal change. Living arrangements encompass a series
of interlocking concepts, including family and household, and frequently form the
basis of data collection, analyses and theorising. Broadly speaking, there are two
linked questions. What are the types of and changes in living arrangements? What
causes living arrangement change and variation?
The meanings associated with households and families continue to change, both
at the individual level and at the normative or societal level. Such change is observed
throughout history (Gillis, 2004; Jamieson et al., 2002; Elizabeth, 2000; Scott,
1999; Manting, 1996). In part changes in meanings, attitudes, values and beliefs
are a function of generational change (Manning et al., 2007; Hall, 2006; Axinn
and Thornton, 2000; Lewis, 1999, 2001b). Normative views on living arrange-
ments, especially marriage, have always shifted (Coontz, 2004; Smock and Man-
ning, 2004; Thornton et al., 2007). For example, there is greater acceptance of non-
marital relationships (Thornton and Young-DeMarco, 2001), explained in part by
greater experience of new forms of living arrangements by greater proportions of
the population as a whole. Barlow (2005), however, makes the important distinc-
tion between accepting and tolerating new(er) forms of living arrangements, such
as non-marital cohabitation and parenting, and argues that acceptance is replacing
tolerance.
Much effort – both academic and political – has been expended into better
understanding the decline in ‘traditional’ family arrangements and associated chal-
lenges to our understanding of the ways in which people live together. At the heart
of this endeavour is a better understanding of relationships and living arrange-
ments. The challenge is not only to capture and describe these trends in living
arrangements, but also to better understand the processes that explain this change
(Seltzer et al., 2005). It is worth considering what is meant by this traditional fam-
ily, not least because its construction is time and space-specific, at its core child-
bearing and rearing and sexual intimacy. It might be perceived as involving notions
of social and legal recognition, combined with concepts of obligations and rights
for the couple, all of which are rapidly shifting in the western world, mediated by
gender, ethnicity (MacLean and Eekelaar, 2004), culture and religion (Eekelaar and
Maclean, 2004; Lehrer, 2004). The picture is further complicated by heterogeneity
across (Heuveline and Timberlake, 2004; Kiernan, 2001; Raley, 2001; Seltzer, 2004;
Wagner and Weib, 2006) and within countries (Liefbroer and Dourleijn, 2006). In
sum, across a range of settings, men and women might be described as being less
dependent on marriage and the family for the fulfilment of a range of needs, includ-
ing nurture, companionship and happiness.
8 D. Kneale et al.
Perspectives about whether change is beneficial or otherwise to society can

be highly polarised. Ranging from constructs of the selfish individual (Morgan,
2000) to an outcome of the pursuit of democratic and consensual relationships
(Giddens, 1993). Interestingly, and perhaps unsurprisingly, the latter perspective
is most likely to be reported by cohabitees (Lewis, 2001a) and unmarried young
people (White, 2003).
Legal systems have grappled for decades with how to best accommodate the mul-
tiple and changing forms of living arrangements (Probert, 2004; Therborn, 2007).
Across Europe the development of statutory regulation of non-marital cohabitation
has begun with, for example, French PACS and Dutch ‘Registered Partnerships’
(Bradley, 2001). Theorising about the causes driving these changes in living arrange-
ments covers a broad spectrum, underpinned by demographic change, most notably
declining fertility and population ageing. Theorists have argued that living arrange-
ment changes are a response to, and at times a cause of, processes of individualisa-
tion (Alders and Manting, 2001), secularisation (Lesthaeghe and Moors, 1995) and
risk identification and avoidance.
Living arrangements are becoming increasingly diverse (Allan et al., 2001),
complex (ESRC, 2006), and multi-directional. They include: a rise in post-marital
cohabitation relative to higher order marriages; reconciliations and multiple
separations (Binstock and Thornton, 2003); multiple union creation and dis-
solution, described as ‘sequential marital monogamy’ (De Graaf and Kalmijn,
2003); non-co-resident step-parenting relationships and childrearing (Ermisch
and Francesconi, 2000; Bumpass et al., 1995); same sex unions; complex
carer relationships (familial and commercial); a proliferation of childrearing
arrangements (Seltzer, 2000); growing rates of non-marital relationships for
older populations post-marriage or bereavement (De Jong Gierveld, 2004;
Mahay and Lewin, 2007); and, living-apart-together (LAT) relationships (Levin,
2004). Processes of globalisation and population mobility further add to the
heterogeneity of living arrangements. Bledsoe (2006) has identified, for exam-
ple, new forms of family creation among African migrant communities in
Europe.
In moving from the perceived ‘traditional’ to the contemporary, as some of the
emergent forms of living arrangement imply, there are complex inter-relationships
above and beyond the dyad. Indeed, dyad relationship formation can result in many
different forms of family structure. These relationships extend above and beyond
simply who co-resides with whom. For example, Eggebeen (2005) finds that dyad
relationship type was significantly associated with levels of support provided to par-
ents. Cohabiting young adults were significantly less likely to exchange support
with their parents than their married or single counterparts. Trends and processes
in living arrangements do not operate in a vacuum from other processes of social
change. There are complex inter-relationships, for example, between union transi-
tions and other major life course transitions, including: (un)employment, education,
geographic mobility, property ownership, fertility (Berrington and Diamond, 2000;
Haskey, 2001; Oppenheimer, 2003; Flowerdew and Hamad, 2004; Osborne, 2005;
Guzzo, 2006; Lauster, 2006; Musick, 2007).
This very complexity has implications for how we study, and the data we use to
study, peoples’ living arrangements. The processes of dynamics in living arrange-
ments continue to be less well understood than the trends, which tend to be more
amenable to secondary analysis of quantitative datasets. If we are to better under-
stand the factors that affect changes in living arrangements, then we need to know
what these living arrangements mean to those involved, whether they are a childless
couple, an elderly parent and their middle-aged child, or a complex step-family with
non-co-residential children. How do people respond to survey-based questions and
categorise themselves and what might the implications be for analyses of contempo-
rary living arrangements? (Glaser et al., 2005; Hunter, 2005; Knab and McLanahan,
2006; Murphy, 2000). In Chapter 6, Ernestina Coast uses prospective data from the
British Household Panel Survey to analyse individuals’ relationship expectations
and subsequent outcomes between 1998 and 2005 and to investigate how attitudes
towards cohabitation differ by age, sex, previous relationship history and parent-
hood. Her analyses underscore the heterogeneity of cohabitation and its meaning
for cohabiters. The data shed some light on why people might cohabit, and for
never-married respondents, suggest that cohabitation represents a way of assessing
partner compatibility. Cohort changes in the experience of, and attitudes towards,
cohabitation, emphasise underlying and powerful normative changes in society due
to intra-generational change and generation succession.
As social scientists we need to consider data, and the way in which we collect
and use them to understand the processes at work behind changing living arrange-
ments. To illustrate this point, two examples are drawn from the body of evidence
for non-marital cohabitation. De Vaus et al. (2005) note that the timing of evidence
is crucial for our understanding of demographic processes. Much of the data used to
theorise about, for example, the influence of pre-marital cohabitation on subsequent
divorce, has been based on evidence from couples who cohabited in the 1970s and
1980s when cohabitation was much less commonplace in the general population.
Secondly, living arrangements can be increasingly commonplace in society whilst
being ‘statistically invisible’, viz cohabitation prior to the 1970s (Kiernan, 2000).
Mixed methods and qualitative approaches to studying living arrangements
are relatively under-developed compared to quantitative approaches (Lewis,
2001a; Lampard and Peggs, 1999). There is an emerging body of qualitative
research into the meanings of living arrangements (MacLean and Eekelaar, 2004;
Manning and Smock, 2005), with specific focus on the inter-relationship between
parenthood and union type (Reed, 2006; Gibson-Davis et al., 2005; Sassler, 2004;
Smock et al., 2005).
Establishing good living arrangements is of primary importance in creating a
good life style for the individuals involved but living arrangements may also be
responsible for negative attributes such as loneliness, stress and intolerance that
may lead to ill-health and unhappiness. In fact, living arrangements are closely
associated with multiple aspects of well-being (Stafford et al., 2004; Dush and
Amato, 2005), happiness (Zimmermann and Easterlin, 2006), risk behaviours
(Duncan et al., 2006), domestic violence (Kenney and McLanahan, 2006) and men-
tal health (Marcussen, 2005; Mastekaasa, 2006). There are two chapters in this
10 D. Kneale et al.
volume that are concerned with exploring the relationship between living arrange-
ments and health and well-being of particular groups. In Chapter 7, Harriet Young
and Emily Grundy focus on the possible consequences of different types of liv-
ing arrangements for the health and well-being of older people. Using data from
longitudinal studies, they show that older people living with a spouse had the
highest levels of health and well-being in England and Wales, except for older
women living alone who rated their health as better than those living with a
spouse. Among the unmarried, on the other hand, those living alone considered
themselves healthier than those living with others but more likely to be depressed
and lonely than those living with others. They found some interesting variations
in these associations across Europe, due to differences in culture and welfare
regimes.
One increasingly common form of living arrangement is that associated with
stepparenting and in Chapter 8, Peteke Feijten, Paul Boyle, Zhiqiang Feng, Vernon
Gayle and Elspeth Graham report on their study that attempts to assess the impact
that stepparenthood has on the mental health of stepparents or their partners. In this
case, they use another longitudinal study, the National Child Development Study
(NCDS) to investigate a series of hypotheses which suggest adults living in step-
families have a higher risk of having poor mental health than comparable adults in
conventional families, although this effect may partly be due to selection of respon-
dents with prior mental health problems into stepfamilies.
The household as a unit of analysis might be perceived as becoming less complex
and smaller, not least through the rise in single person households and stepparent-
ing arrangements. However, this means that we will need to shift our focus away
from household-based sources of information and analyses, and acknowledge the
growing importance of non-co-residential rights, obligations and networks. This is
relevant across all stages of the lifecycle, and is becoming increasingly important at
older ages.
Care
There are significant care implications arising from the research on living arrange-
ments for older people reported in Chapter 7 and on stepparenting presented in
Chapter 8. On a more general global level, demographic ageing poses huge chal-
lenges for societies since it will affect pension and social security systems, health
care provision and the needs of both dependent children and particularly, the infirm
elderly for family, social and state care. The medium variant of the United Nations
world population projections (United Nations, 2005) indicates that in western
and central Europe, the so-called EU25+ (25 EU members plus another 3 EEA
members plus Switzerland), the size of the working age population (age 15–64)
which in 2005 was 317 million, will start to decline after 2015 reaching 302 mil-
lions in 2025 and 261 million in 2050, a decline of 18%. On the other hand, due
to increasing life expectancy and the ageing of the baby boom generation, the 65+
Table 1.1 Old age dependency ratio, 2005–2050, selected countries
Percentage change
Country 2005 2025 2050 2005–2050
Belgium 26.3 36.5 48.1 82.9

Czech Republic 19.8 35.0 54.8 176.8
Finland 23.7 41.4 46.7 97.0
France 25.3 36.9 47.9 89.3
Germany 27.8 39.3 55.8 100.7
Ireland 16.5 25.2 45.3 174.5
Italy 29.4 39.7 66.0 124.5
Slovakia 16.3 28.1 50.6 210.4
Sweden 26.4 36.5 40.9 54.9
United Kingdom 24.4 33.2 45.3 85.9
EU 25 average 24.9 35.7 52.8 112.0
Source: Eurostat (2004) based on the assumption that net immigration

will amount to almost 40 million between 2005 and 2050.
age group will grow from 79 million in 2005 to 133 million in 2050, an increase
of 68%, with the largest increases occurring for those people over 80 years of age.
These figures are staggering; the old age dependency ratio in EU25 which, in 2005,
was approximately 25 people aged over 65 to every 100 in the working age range,
will more than double to almost 53 people in the age group 65+ per 100 of working
age. Dependency ratios for a selection of countries in EU25 from Eurostat (2004)
illustrate the extent of the challenge (Table 1.1). Whilst the UK has to consider a
lower than average 86% increase in the dependency ratio, Ireland and the Czech
Republic are both set to experience an increase of around 175% and Slovakia’s pro-
jected increase exceeds 200%.
These changes will have a profound influence not only on the demand for care
for the elderly but also on the complete state of intergenerational relations. We
should not forget that Britain’s welfare state was founded on an implicit intergen-
erational contract based on a principle of reciprocity, such that each generation,
during its productive years, supports both younger and older generations in antic-
ipation that when reaching a time of dependency itself, it can expect to receive
support from subsequent generations. It is a contract that has been characterised as
being based on duty, national collectivity and intergenerational solidarity (Walker,
1996; Phillipson, 1998). This intergenerational contract is already under pressure
due to changing social attitudes rather than numbers. In Chapter 7, Harriet Young
and Emily Grundy identify substantial changes in the living arrangements of older
people, who are now more likely to live alone and less likely to live with rela-
tives in multi-generational households. Low fertility and low mortality are altering
intergenerational patterns within families such that it will become more and more
common for families to have two generations of retirees, putting an increased bur-
den of care on the middle ‘productive’ or ‘pivot’ generation who, at the same time,
may experience delayed or indeed loss of inherited wealth due to their parents’ care
needs (Bengston et al., 1991).
12 D. Kneale et al.
As the generational contract has also been predicated on a gender contract (based
on men’s economic and women’s caring contributions), the increased duration and
intensity of caring activities will impact particularly on women but also increas-
ingly on men. Women are playing an increasingly significant role in the labour
market without any enhanced provision of state childcare, at a time when other
caring demands are increasing, and men are becoming significant contributors of
unpaid caring labour too (Buckner and Yeandle, 2006, 2007), although men’s demo-
graphic behaviour has received little attention relative to women. These develop-
ments raise a set of challenges about how the work/care conundrum can be resolved
for the productive generation; how organisational cultures/structures might need to
change to accommodate this; and how gender and caring roles and relations might
be transformed in the process (Williams, 2004; Yeandle, 2007). In Chapter 9, Ali-
son Smith Koslowski’s review of current literature confirms that in most western
European countries, grandparents have become a very important source of child-
care, particularly for infants. Her review highlights the gap in knowledge (and data)
relating to informal grandparental care for grandchildren. She demonstrates the pos-
sible contradictions of caring for grandchildren versus a range of policies associated
with ‘active ageing’, itself an underspecified term. This work, based on analyses
of a range of datasets, demonstrates the need to reconsider our understandings of
grandparenthood, its contributions and meanings.
There are major questions about how care will be provided in the context of ris-
ing life expectancy, divorce rates and higher dependency ratios. Historically, unpaid
and family care for sick, frail or disabled family members, and dependent children,
usually delivered in the home, was provided mainly by women. However, the ero-
sion of the ‘male breadwinner model’, together with other changes, means that more
women are active in the paid labour force – while men have begun to be drawn into
family caring roles in larger numbers, especially in middle and later life in support
of very aged parents, or of sick or disabled partners; nevertheless, gendered assump-
tions about responsibility for care, within and outside the family, remain strong and
persistent.
The 2001 Census included, for the first time, a question on the provision of
unpaid care: ‘Do you look after or give any help or support to family members,
friends or neighbours or others because of: long-term physical or mental ill-health
or disability or problems related to old age?’. This revealed that, across England
and Wales, 10% of the population – almost 5.2 million people – provide unpaid
care, and almost 3.9 million carers are of working age of whom 1.5 million combine
full-time paid employment with unpaid care. Of these working carers, 58% are men
(Buckner and Yeandle, 2006). Moreover, the longer lives of disabled children and
the increased longevity of sick and older people mean unpaid caring roles can last
for many years – sometimes for decades. Shortages of labour in health and social
care already pose problems for the delivery of formal care services, where recruiting
and retaining staff and expanding the pool of potential recruits has proved very
challenging in recent decades (Yeandle et al., 2006a, b). Most older people express
a preference for independence and care at home and hospital discharge policies
promote additional domiciliary care. Yet, the traditional source of domiciliary care
workers (unqualified, middle aged female returned to the labour market) is shrink-
ing fast. Migrant workers are often cited as a source of extra caring labour but how
sustainable this is in the longer term is open to question (Ungerson and Yeandle
2007). Migration brings the challenges of transnational care, with carers living and
caring in different countries.
Mobility
International migration has received considerable attention in the academic and

policy-oriented literature (e.g. Tamas and Münz, 2006). Some scholars have
reviewed migration dynamics in various world regions (Appleyard, 1998); others
have focused on theories within particular disciplines (Massey et al., 1998; Poot
et al., 1998). Skeldon (1997) draws attention to the lack of clarity in the linkages
between migration and poverty whilst a host of commentators have considered the
effects of immigration on destination countries (such as Dustmann et al., 2003;
Borjas, 2004), the effects of emigration on source countries (such as Fischer et al.,
1997; Collyer, 2004), the role of remittances (such as Terry and Wilson, 2005),
migration and the brain drain (such as Kapur and McHale, 2005), and the role of
diasporas in development (such as Levitt, 2006).
Whilst international migration tends to grab the headlines, we should recog-
nise that the flows of immigrants are relatively small compared with the volume
of movement taking place within the UK and permanent changes of usual residence
are themselves relatively small compared with the level of daily mobility, much of
which is associated with the journey to work or to study. The 2001 Census tells
us that just under 400,000 immigrants arrived in Great Britain in the 12 month
period before the Census compared with over 6 million people moving internally
out of the total population of 57.1 million. These are underestimates because we
know that a further 450,000 people migrated but the Census has no record of their
usual address at the start of the period. In contrast to these volumes, consider the
daily journey to work in Great Britain involving 25.7 million trips of those of work-
ing age. These selected statistics, extracted online using the Web-based Interface
to Interaction Data (WICID) (Stillwell and Duke-Williams, 2003), exemplify the
extent of migration and commuting but tell us nothing about all the other interac-
tion behaviour that we engage in. Census data underpins much of the migration and
commuting research in the UK (such as Champion, 2005; Dennett and Stillwell,
2008; Frost and Shepherd, 2004; Coombes and Raybould, 2001), focusing in most
cases on flows of individual migrants or commuters.
There are important associations between migration, commuting and living
arrangements and in Chapter 10, Oliver Duke-Williams attempts to look more
closely at the different units of migration. The household has always been a key
unit in the decision-making process relevant to residential mobility (Rossi, 1955).
Whilst many households consist of singletons, the decision to move for families
is frequently the result of a combination of factors – many related to the life
14 D. Kneale et al.
Fig. 1.2 Proportions of 90

1 person 2 persons 3+ persons
migrants moving as wholly 80
moving households or other
groups by type of household, 70
2000–01 60
Percentage
(Source: Census 2001 Special 50
Migration Statistics)
40
30
20
10
0
Groups Migrants Groups Migrants
Wholly moving Other moving
households groups
cycle – impacting on different members of the group. Precise numbers of individu-

als involved in moving usual residence are available from successive censuses but
the use of moving groups in the 2001 Census enabled some insights to be gained
into the relative proportions of those moving together. Figure 1.2 illustrates the
proportions of groups and migrants in one, two and three or more person house-
holds moving internally in the United Kingdom in the year before the Census either
in wholly moving households and in other moving groups. Amongst the wholly
moving households, single person groups dominate but over half the migrants are
moving as families in households of three or more persons. Single migrants domi-
nate the other moving groups to a much greater extent with a similar proportion of
migrants in the 2 person and 3+ person households.
One key limitation with UK Census migration data is the absence of any question
about motivation and yet it is this attribute which has captured the attention of migra-
tion analysts for many years seeking to distinguish those who move for economic
or job reasons from those who keep the same job but change house, for example.
Oliver Duke-Williams turns to the British Household Panel Survey in Chapter 10 to
find answers to questions about motivation and demonstrates how motivations vary
according to the type of household under consideration. In the case of lone parents,
the move is motivated by the split from a partner, whereas single elderly people tend
to move for health reasons.
In a recent audit of interaction data, Dennett et al. (2007) have drawn attention
to the need for more research on different types of mobility using secondary data
sets that exist but have not been used hitherto to their full capacity. These data
sets include those based on administrative records and involve the collection of
records arising from some transaction, registration or as a record of service delivery.
They are collected for administrative rather than purely research purposes and
many of these data sets are collected by Government departments (Jones and Elias,
2006). These data sources include variables that provide information about either
the migration or the commuting characteristics of NHS patients, school pupils,
university students, asylum seekers, new migrant workers or those attending hos-
pital. In some cases, registration data have much simpler structure than census data
and are only available at a relatively aggregate spatial scale but are particularly valu-
able because they are produced on a regular temporal basis. In other cases, the infor-
mation on migration or mobility has to be generated from the primary unit data using
time-consuming data matching and manipulation algorithms.
One of these relatively new and unexplored administrative data sets, the Pupil
Level Annual School Census (PLASC), is the focus for Joan Wilson’s research
reported in the last chapter of the book. Whilst the Census in Scotland provides
details of the daily travel to study for students and children, similar data are not
produced for England and Wales or Northern Ireland, However, the PLASC does
collect data from each education authority in England and Wales on the location of
pupils and the schools that they attend, potentially providing an extremely useful
data set on the journey to school. Various data sets are collected and held by the
Department for Education and Skills (DfES) within a centralised ‘data warehouse’,
including the National Pupil Database (NPD), local authority data, school level data,
school workforce data and geographical data (Ewens, 2005; Jones and Elias, 2006).
The NPD was established in 2002 and contains linked individual pupil records for
all children in the state school system which is updated annually. Each pupil is given
a unique pupil number (UPN) and has an associated set of attributes: age, gender,
ethnicity, special educational needs, free school meal entitlement, key stage assess-
ments, public exam results, home postcode and school attended. It is the availability
of the last two attributes which gives the possibility of identifying various mobility
characteristics.
The linking of pupils from one year to the next using the UPN means that a
longitudinal profile of each pupil is available whose extent depends on how long
the pupil has been in the education system. Potentially, this means that pupils can
be tracked over time and their transitions through the education system can be
identified, including their movements between schools and between different home
addresses (Harland and Stillwell, 2007a, b). The PLASC is therefore a potential
source of data on commuting to school, i.e. on child migration from one usual resi-
dence to another and on pupil mobility between schools.
It is perhaps not surprising that these introductory comments on the content of
the book which also attempt to provide some context to what follows, have come
to an end with a discussion about data. The Understanding Population Trends and
Processes programme, is, after all, about the analysis of secondary data sets. So,
in completing this chapter, we provide a summary of the data sources that the con-
tributors to this volume have used. Table 1.2 indicates that a variety of different
longitudinal and cross-sectional survey and census data sets have been employed
in the analyses that are reported in the book. No further explanation of these
data sources is attempted here since each will be introduced in the corresponding
chapter, together with the methods that have been adopted for data extraction, prepa-
ration and analysis as appropriate.
In drawing this chapter to a close, we hope that readers will find the chapters
in this volume valuable as a means of understanding more about the current
16 D. Kneale et al.
Table 1.2 Main data sets used in forthcoming chapters
Chapter Author(s) Data sources
2 Simpson National Child Development Study

(NCDS); British Cohort Study
(BCS70)
3 Smith and Ratcliffe British Household Panel Survey
(BSPS); Family Expenditure
Survey (FES); Family
Resources Survey (FRS)
4 Hansen et al. British Birth Cohort Study
(BCS70); Millennium Cohort
Study (MCS); National Child
Development Study (NCDS)
5 Kneale National Child Development Study
(NCDS); British Cohort Study
(BCS70)
6 Coast British Household Panel Survey
(BHPS)
7 Young and Grundy English Longitudinal Study of
Ageing (ELSA); Office for
National Statistics (ONS)
Longitudinal Study (LS);
European Social Survey (ESS)
8 Boyle et al. National Child Development Study
(NCDS)
9 Smith Koslowski Survey of the European
Community Household Panel
(ECHP); the European Social
Survey (ESS); Growing Up in
Scotland (GUS); Millennium
Cohort Study (MCS)
10 Duke-Williams 2001 Census Origin-Destination
Statistics; British Household
Panel Study (BHPS)
11 Wilson Pupil Level Annual School Census
(PLASC)
patterns and processes of fertility, living arrangements, care and mobility, but also
for helping to recognise what implications these trends and processes have for prac-
titioners and policy makers. Both of these are primary aims of the UPTAP initiative,
along with the promotion of the use of large-scale social science data sets and
building capacity in secondary data analysis amongst new and mid-career
researchers. Further volumes in the series will extend our understanding by
covering social and spatial disparities (Volume 2) and ethnicity and integration
(Volume 3).
References
Alders, M.P.C. and Manting, D. (2001) Household scenarios for the European Union, 1995–2025,
Genus, LVII(2): 17–48.
Allan, G., Hawker, S. and Crow, G. (2001) Family diversity and change in Britain and Western
Europe, Journal of Family Issues, 22(7): 819–837.
Appleyard, R. (1998) Emigration Dynamics in Developing Countries Volume 1–3 Sub-Saharan
Africa; Mexico, Central America and the Caribbean; South Asia and the Arab Region, UNFPA
and IOM, Ashgate, Aldershot.
Axinn, W. G. and Thornton, A. (2000) The Transformation in the Meaning of Marriage, in Waite,
L., Bachrach, C., Hindin, M., Thomson, E. and Thornton, A. (eds.) Ties That Bind: Perspectives
on Marriage and Cohabitation, Aldine de Gruyter, New York, pp. 147–165.
Barlow, A. (ed.) (2005) Cohabitation, Marriage and the Law: Social Change and Legal Reform in
the 21st Century, Oxford, Portland OR, Hart Publishing.
Bengston, V., Marti, G. and Roberts, R. (1991) Age group relations: generational equity and
inequity, in Pillemer, K. and McCartney, K (eds.) Parent-Child Relations Across the Lifespan,
Lawrence Erlbaum, Hillside (NJ), pp. 253–278.
Berrington, A.M. (2004) Perpetual postponers? Women’s, men’s and couple’s fertility intentions
and subsequent fertility behaviour, Population Trends, 117: 9–19.
Berrington, A. and Diamond, I. (2000) Marriage or cohabitation: a competing risks analysis of
first-partnership formation among the 1958 British birth cohort, Journal of the Royal Statistical
Society, 163(2): 127–152.
Binstock, G. and Thornton, A. (2003) Separations, reconciliations, and living apart in cohabiting
and marital unions, Journal of Marriage and Family, 65(2): 432–443.
Bledsoe, C.E. (2006) The demography of family reunification: from circulation to substitution in
Gambian Spain, MPIDR Working Paper No 2006–053, Max Planck Institute for Demographic
Research, Rostock.
Borjas, G. (2004) Increasing the supply of labour through immigration: measuring the impact of
native-born workers, CES Working Paper No 504, www.cis.org.
Bradley, D. (2001) Regulation of unmarried cohabitation in west-European jurisdictions
– determinants of legal policy, International Journal of Law, Policy and the Family,
15(1): 22–50.
Buckner, L. and Yeandle, S. (2006) More than a Job Working Carers: Evidence from the
2001 Census, Carers UK, London. Available from: http://www.carersuk.org/ Newsandcam-
paigns/News/1159285953/MorethanaJob.pdf.
Buckner, L. and Yeandle, S. (2007) Care and Caring in EU Member States, Carers UK, London.
Bumpass, L.L., Raley, R.K. and Sweet, J.A. (1995) The changing character of stepfamilies: impli-
cations of cohabitation and nonmarital childbearing, Demography, 32(3): 425–436.
Champion, A.G. (2005) Population movement within the UK, in Chappell, R. (ed.) Focus on People
and Migration 2005 Edition, Palgrave Macmillan, Basingstoke, pp. 91–113.
Collyer, M. (2004) The development impact of temporary international labour migration on
southern Mediterranean sending countries, Working Paper T6, Sussex Centre for Migration
Research, University of Susses, Brighton.
Coombes, M.G. and Raybould, S. (2001) Commuting in England and Wales: ‘people’ and ‘place’
factors, European Research in Regional Science, 11(1): 111–133.
Coontz, S. (2004) The world historical transformation of marriage, Journal of Marriage and Fam-
ily, 66(4): 974–979.
Coyle, A. (2005) Changing times: flexibilization and the re-organization of work in feminized
labour markets, Sociological Review, 53: 73–88.
Darroch, J.E., Landry, D.J. and Oslak, S. (1999) Pregnancy rates among US women and their
partners in 1994, Family Planning Perspectives, 31: 122–126+136.
De Graaf, P.M. and Kalmijn, M. (2003) Alternative routes in the remarriage market: competing-risk
analyses of union formation after divorce, Social Forces, 81(4): 1459–1598.
18 D. Kneale et al.
De Jong Gierveld, J. (2004) Remarriage, unmarried cohabitation, living apart together: part-
ner relationships following bereavement or divorce, Journal of Marriage and Family, 66(1):
236–243.
De Vaus, D, Qu, L. and Weston, R. (2005) The disappearing link between premarital cohabitation
and subsequent marital stability, 1970–2001, Journal of Population Research, 22(2): 99–118.
Dennett, A. and Stillwell, J. (2008) Internal migration in Great Britain – a district level analysis
using 2001 Census data, Working Paper 01/08, School of Geography, University of Leeds,
Leeds.
Dennett, A., Stillwell, J. and Duke-Williams, O. (2007) Interaction data sets in the UK: an audit,
Working Paper 07/03, School of Geography, University of Leeds, Leeds.
Dex, S, Joshi, H, Macaran, S. and McCulloch, A. (1998) Employment after childbearing and
women s subsequent labour force participation: evidence from the British 1958 birth cohort,
Oxford Bulletin of Economics and Statistics, 60: 79–98.
Duncan, G. J, Wilkerson, B. and England P. (2006) Cleaning up their act: The effects of marriage
and cohabitation on licit and illicit drug use, Demography, 43(4): 691–710.
Dush, C.M.K. and Amato, P.R. (2005) Consequences of relationship status and quality for subjec-
tive well-being, Journal of Social and Personal Relationships, 22(5): 607–628.
Dustmann, C, Fabbri, F, Preston, I. and Wadsworth, J. (2003) The Local Labour Market Effects of
Immigration in the UK, Home Office Online Report 06/03, London.
Edwards, M.E. (2002) Education and occupations: reexamining the conventional wisdom about
later first births among American mothers, Sociological Forum, 17: 423–443.
Eekelaar, J. and Maclean, M. (2004) The obligations and expectations of couples within families:
three modes of interaction, Journal of Social Welfare and Family Law, 26(2): 117–130.
Eggebeen, D.J. (2005) Cohabitation and exchanges of support, Social Forces, 83(3): 1097–1110.
Elizabeth, V. (2000) Cohabitation, marriage, and the unruly consequences of difference, Gender
and Society, 14(1): 43–56.
Elliott, J. (2005) Using Narrative in Social Research, SAGE Publications Limited, London.
Ermisch, J. and Francesconi, M. (2000) The increasing complexity of family relationships: life-
time experience of lone motherhood and stepfamilies in Great Britain, European Journal of
Population, 16(3): 235–250.
ESRC (2006) Changing Household and Family Structures and Complex Living Arrangements,
Economic and Social Research Council. Swindon.
Eurostat (2004) EUROPOP 2004 – Summary Note on Assumptions and Methodology for Inter-
national Migration, Working Paper for the Ageing Working Group on the EPC, ESTAT/F1 -
/POP/19(2004)/GL, Luxembourg.
Ewens, D. (2005) The national and London pupil datasets: an introductory briefing for researchers
and research users, Data Management and Analysis Group, Greater London Authority.
Fischer, P, Martin, R. and Straubhaar, T. (1997) Interdependencies between development and
migration, in Hammar, T., Brochmann, G., Tamas, K. and Faist, T. (eds.) International
Migration, Immobility and Development: Multidisciplinary Perspectives, Berg Publishers,
Oxford.
Flood, M. (2007) Involving men in gender policy and practice, Critical Half: Bi-Annual Journal of
Women for Women International, 5: 9–15.
Flowerdew, R. and Hamad, A. (2004) The relationship between marriage, divorce and migration
in a British data set, Journal of Ethnic and Migration Studies, 30(2): 339–351.
Frost, M. and Shepherd, J. (2004) New methods of assessing service provision in rural
England, in Stillwell, J. and Clarke, G. (eds.) Applied GIS and Spatial Analysis, Wiley,
Chichester.
Gibson-Davis, C.M, Edin, K. and McLanahan, S. (2005) High hopes but even higher expectations:
the retreat from marriage among low-income couples, Journal of Marriage and the Family,
67(5): 1301–1312.
Giddens, A. J. (1993) The Transformation of Intimacy: Love, Sexuality and Eroticism in Modern
Societies, Polity Press, Cambridge.
Gillis, J. (2004) Marriages of the mind, Journal of Marriage and Family, 66(4): 988–991.
Glaser, K, Askahm, J. et al. (2005) Married or single: which shall I tick? Findings from a Study of
BHPS Marital Status Data, BHPS 2005 Conference, Colchester.
Gonzalez, M.-J. and Jurado-Guerrero, T. (2006) Remaining childless in affluent economies: a com-
parison of France, West Germany, Italy and Spain, 1994–2004, European Journal of Popula-
tion, 22: 317–352.
Goodman, A, Kaplan, G. and Walker, I. (2004) Understanding the effects of early motherhood
in Britain: the effects on mothers, IFS Working Paper 04/18, Institute for Fiscal Studies,
London.
Greene, M.E. and Biddlecom, A.E. (2000) Absent and problematic men: demographic accounts of
male reproductive histories, Population and Development Review, 26: 81–115.
Guzzo, K.B. (2006) The relationship between life course events and union formation, Social Sci-
ence Research, 35(2): 384–408.
Hadfield, L., Rudoe, N. and Sanderson-Mann, J. (2007) Motherhood, choice and the British media:
a time to reflect, Gender and Education, 19: 255–263.
Hall, S.S. (2006) Marital meaning: exploring young adults’ belief systems about marriage, Journal
of Family Issues, 27(10): 1437–1458.
Harden, A., Brunton, G., Fletcher, A., Oakley, A., Burchett, H. and Backhans, M. (2006) Young
people, pregnancy and social exclusion: A systematic synthesis of research evidence to iden-
tify effective, appropriate and promising approaches for prevention and support, The Evidence
for Policy and Practice Information and Co-ordinating Centre (EPPI-Centre), University of
London, London.
Harland, K. and Stillwell, J. (2007a) Commuting to school in Leeds: how useful is the PLASC?
Working Paper 07/02, School of Geography, University of Leeds, Leeds.
Harland, K. and Stillwell, J. (2007b) Using PLASC data to identify patterns of commuting to
school, residential migration and movement between schools in Leeds, Working Paper 07/03,
School of Geography, University of Leeds, Leeds.
Haskey, J. (2001) Demographic aspects of cohabitation in Great Britain, International Journal of
Law, Policy, and the Family, 15(1): 51–67.
Heuveline, P. and Timberlake, J.M. (2004) The role of cohabitation in family formation:
the United States in comparative perspective, Journal of Marriage and Family, 66(5):
1214–1230.
Holzmann, R. and Münz, R. (2004) Challenges and Opportunities of International Migration for
the EU, Its Member States, Neighbouring Countries and Regions: A Policy Note, World Bank,
Washington DC and Institute for Futures Studies, Stockholm.
Hotz, V. J., MCelroy Williams, S. and Sanders, S.G. (2004) Teenage Childbearing and Its Lifecy-
cle Consequences: Exploiting a Natural Experiment, Department of Economics, University of
California (UCLA), Los Angeles.
Hunter, J. (2005). Report on Cognitive Testing of Cohabitation Questions. Study Series Report:
Survey Methodology #2005-06, U.S. Census Bureau, Statistical Research Division: 18.
Imamura, M., Tucker, J., Hannaford, P., Oliveira Da Silva, M., Astin, M., Wyness, L.,
Bloemenkamp, K.W.M., Jahn, A., Karro, H., Olsen, J. and Temmermen, M. (2007) Factors
associated with teenage pregnancy in the European Union countries: a systematic review, Euro-
pean Journal of Public Health, 17: 630–636.
Jamieson, L., Anderson, M., McCrone, D., Bechhofer, F., Stewart, R., and Li, Y (2002) Cohabita-
tion and commitment: partnership plans of young men and women, Sociological Review, 50(3):
356–377.
Jefferies, J. (2005) The UK Population: Past, Present and Future. Focus on People and Migration,
Office for National Statistics, London.
Jones and Elias, P. (2006) Administrative data as research resources: a selected audit, Working
Paper, Warwick Institute for Employment Research, Warwick.
Joshi, H. (2002) Production, reproduction, and education: women, children and work in a British
perspective, Population and Development Review, 28: 445–474.
20 D. Kneale et al.
Joshi, H. (2007) Social Polarisation in Reproduction. From Boom to Bust: Fertility, Ageing and
Demographic Change, CentreForum, London.
Kapur, D. and McHale, J. (2005) Give Us Your Best and Brightest: the Global Hunt for talent and
its Impact on the Developing World, Center for Global Development, Washington DC.
Kenney, C.T. and McLanahan, S.S. (2006) Why are cohabiting relationships more violent than
marriages? Demography, 43(1): 127–140.
Kiernan, K. (2000) European perspectives on union formation, in Waite, L. (ed.) The Ties that
Bind: Perspectives on Marriage and Cohabitation, Aldine de Gruyter, New York, pp. 40–58.
Kiernan, K. (2001) The rise of cohabitation and childbearing outside marriage in western Europe,
International Journal of Law, Policy and the Family, 15(1): 1–21.
Knab, J. and McLanahan, S. (2006) Measuring cohabitation: does how, when and who you ask
matter? in Hofferth, S.L. and Casper, L.M. (eds.) Handbook of Measurement Issues in Family
Research, Lawrence Erlbaum & Associates, New Jersey, pp. 19–33.
Lampard, R. and Peggs, K. (1999) Repartnering: the relevance of parenthood and gender to
cohabitation and remarriage among the formerly married, British Journal of Sociology, 50(3):
443–465.
Lauster, N.T. (2006) A room of one s own or room enough for two? Access to housing and new
household formation in Sweden, 1968–1992, Population Research and Policy Review. 25(4):
329–351.
Lehrer, E.L. (2004) The role of religion in union formation: an economic perspective, Population
Research and Policy Review, 23(2): 161–185.
Lesthaeghe, R. and Moors, G. (1995) Living arrangements and parenthood: do values matter? in de
Moor, R.A. (ed.) Values in Western Societies, Tilburg University Press, Tilburg, pp. 217–250.
Lesthaeghe, R. and Neels, K. (2002) From the first to the second demographic transition - an
interpretation of the spatial continuity of demographic innovation in France, Belgium and
Switzerland, European Journal of Population, 18(4): 225–260.
Levin (2004) Living apart together: a new family form, Current Sociology, 52(2): 223–240.
Lewis, J. and Campbell, M. (2007) UK work/family balance policies and gender equality
1997–2005, Social Politics, 14: 4–30.
Levitt, P. (2006) Transnational migration: conceptual and policy challenges, in Tamas, K. and
Palme, J. (eds.) Globalising Migration Regimes: New Challenges to Transnational Coopera-
tion, Ashgate, Avebury.
Lewis, J. (1999) Marriage, Cohabitation and the Law: Individualism and Obligation, Lord Chan-
cellor’s Department, London.
Lewis, J. (2001a) Debates and issues regarding marriage and cohabitation in the British and Amer-
ican literature, International Journal of Law, Policy, and the Family, 15(1): 159–184.
Lewis, J. (2001b) The End of Marriage? Individualism and Intimate Relations, E Elgar Pub, Chel-
tenham.
Lewis, J. and Campbell, M. (2007) UK work/family balance policies and gender equality
1997–2005, Social Politics, 14: 4–30.
Liefbroer, A.C. and Dourleijn, E. (2006) Unmarried cohabitation and union stability: test-
ing the role of diffusion using data from 16 European countries, Demography, 43(2):
203–221.
MacLean, M. and Eekelaar J. (2004) The obligations and expectations of couples within families:
three modes of interaction, Journal of Social Welfare and Family Law, 26(2): 117–130.
Mahay, J. and Lewin, A.C. (2007) Age and the desire to marry, Journal of Family Issues, 28(5):
706–723.
Manning, W.D., Longmore, M.A. and Giordano, P.C. (2007) The changing institution of marriage:
adolescents’ expectations to cohabit and to marry, Journal of Marriage and Family, 69(3):
559–575.
Manning, W.D. and Smock, P.J. (2005) Measuring and modeling cohabitation: new perspectives
from qualitative data, Journal of Marriage and Family, 67(4): 989–1002.
Manting, D. (1996) The changing meaning of cohabitation and marriage, European Sociological
Review, 12(1): 53–65.
Marcussen, K. (2005) Explaining differences in mental health between married and cohabiting
individuals, Social Psychology Quarterly, 68(3): 239–257.
Massey, D., Arango, J, Hugo, G., Kouaouci, A., Pelegrino, A. and Taylor, J. (1998) Worlds in
Motion: Understanding International Migration at the End of the Millenium, Clarendon Press,
Oxford.
Mastekaasa, A. (2006) Is marriage/cohabitation beneficial for young people? Some evidence on
psychological distress among Norwegian college students, Journal of Community and Applied
Social Psychology, 16(2): 149–165.
Morgan, P.M. (2000) Marriage-lite: The Rise of Cohabitation and its Consequences, Institute for
the Study of Civil Society, London.
Murphy, M. (2000) Cohabitation in Britain, Journal of the Royal Statistical Society, 163(2):
123–126.
Musick, K. (2007) Cohabitation, nonmarital childbearing, and the marriage process, Demographic
Research, 16(9): 249–286.
Office for National Statistics (2007) Live births: age of mother, 1961 onwards (England and Wales),
Health Statistics Quarterly, 37: 36.
Oppenheimer, V.K. (2003) Cohabiting and marriage during young men s career-development pro-
cess, Demography, 40(1): 127–149.
Osborne, C. (2005) Marriage following the birth of a child among cohabiting and visiting parents,
Journal of Marriage and Family, 67(1): 14–26.
Phillipson, C. (1998) Reconstructing Old Age: New Agendas in Social Theory and Practice, Sage,
London.
Pleck, J.H. (2007) Why could father involvement benefit children? Theoretical perspectives,
Applied Developmental Science, 11: 196–202.
Poot, J., Gorter, C. and Nijkamp, P. (eds.) (1998) Crossing Borders: Regional and Urban Perspec-
tives on International Migration, Ashgate, Aldershot UK.
Power, C. and Elliott, J. (2006) Cohort profile: 1958 British birth cohort (National Child Develop-
ment Study), International Journal of Epidemiology, 35: 34–41.
Probert, R. (2004) Cohabitation in twentieth century England and Wales: law and policy, Law and
Policy, 26(1): 13–32.
Raley, R.K. (2001) Increasing fertility in cohabiting unions: evidence for the second demographic
transition in the United States? Demography, 38(1): 59–66.
Reed, J.M. (2006) Not crossing the “extra line”: how cohabitors with children view their unions,
Rendall, M.S., Clarke, L., Peters, H.E., Ranjit, N. and Verropoulou, G. (1999) Incomplete reporting
of men s fertility in the United States and Britain: a research note, Demography, 36, 135–144.
Rendall, M.S., Couet, C., Lappegard, T., Robert-Bobee, I., Romsen, M. and Smallwood, S. (2005)
First birth by age and education in Britain, France and Norway, Population Trends, 121: 27–34.
Rendall, M.S. and Smallwood, S. (2003) Higher qualifications, first birth timing and further child-
bearing in England and Wales, Population Trends, 111: 18–27.
Rossi, P. (1955) Why Families Move: A Study in the Social Pschologyof Urban Residential Mobil-
ity, Free Press, Glencoe, Illinois.
Sarkadi, A., Kristiansson, R., Oberklaid, F. and Bremberg, S. (2008) Fathers’ involvement and
children’s developmental outcomes: a systematic review of longitudinal studies, Acta Paedia-
tricia, 97: 153–158.
Sassler, S. (2004) The process of entering into cohabiting unions, Journal of Marriage and Family,
66(2): 491–505.
Scott, J. (1999) Family Change: Revolution or Backlash in Attitudes? in McRae, S. (ed.)
Changing Britain: Families and households in the 1990s, Oxford University Press, Oxford,
pp. 68–99.
Seltzer, J.A. (2000) Families formed outside of marriage, Journal of Marriage and Family, 62(4):
1247–1268.
Seltzer, J.A. (2004) Cohabitation in the United States and Britain: demography, kinship, and the
future, Journal of Marriage and Family, 66(4): 921–928.
22 D. Kneale et al.
Seltzer, J. A., Bachrach, C.A. Bianchi, S.M., Bledsoe, C.H., Casper, L.M., Chase-Lansdale, P.L.,
et al. (2005) Explaining family change and variation: challenges for family demographers,
Skeldon, R. (1997) Migration and Development: A Global Perspective, Addison Wells/Longman,
Harlow.
Smallwood, S. and Chamberlain, J. (2005) Replacement fertility: What has it been and what does
it mean? Population Trends, 119: 16–27.
Smock, P. J. and Manning, W.D. (2004) Living together unmarried in the United States: demo-
graphic perspectives and implications for family policy, Law and Policy, 26(1): 87–118.
Smock, P. J., Manning, W.D. and Porter, M. (2005) “Everything’s there except money”: How
money shapes decisions to marry among cohabitors, Journal of Marriage and Family, 67(3):
680–696.
Social Exclusion Unit (1999) Teenage Pregnancy, Social Exclusion Unit, London.
Stafford, L., Kline, S.L., and Rankin, C.T. (2004) Married Individuals, cohabiters, and cohabiters
who marry: a longitudinal study of relational and individual well-being, Journal of Social and
Personal Relationships, 21: 231–248.
Stillwell, J. and Duke-Williams, O. (2003) A new web-based interface to British census of popu-
lation origin-destination statistics, Environment and Planning A, 35: 113–132.
Tamas, K. and Münz, R. (2006) Labour Migrants Unbound? EU Enlargement, Transitional Mea-
sures and Labour Market Effects, Institute for Futures Research, Stockholm.
Terry, D.F. and Wilson, S.R. (eds.) (2005) Beyond Small Change: Making Migrant Remittances
Count, Inter-American Development Bank, Washington DC.
Therborn, G. (2007) Is there a future for the family?, Public Policy Research, 14(1): 41–46.
Thornton, A., Axinn, W.G. and Xie, Y. (2007) Marriage and Cohabitation, University of Chicago
Press, Chicago.
Thornton, A. and Young-DeMarco, L. (2001) Four decades of trends in attitudes toward family
issues in the United States: the 1960s through the 1990s, Journal of Marriage and Family,
63(4): 1009–1037.
Ungerson, C. and Yeandle, S. (eds.) (2007) Cash for Care in Developed Welfare States, Palgrave
Macmillan, London.
UNICEF (2001) A League Table of Teenage Births in Rich Nations, UNICEF Innocenti Research
Centre, Florence.
United Nations (2005) World Population Prospects. The 2004 Revision, UN, NewYork.
Van de Kaa, D.J. (1987) Europe’s second demographic transition, Population Bulletin, 42: 1–59.
Wagner, M. and Weib, B. (2006) On the variation of divorce risks in Europe: findings from a meta-
analysis of European longitudinal studies, European Sociological Review, 22(5): 483–500.
Walker, A. (ed.) (1996) The New Generational Contract: Intergenerational Relations, Old Age and
Welfare, Routledge, London.
White, N.R. (2003) Changing conceptions: young people’s views of partnering and parenting, Jour-
nal of Sociology, 39(2): 149–164.
Williams, F. (2004) Trends in women’s employment, domestic service, and female migration:
changing and competing patterns of solidarity, in Knijn, T. and Komter, A. (eds.) Solidarity
Between the Sexes and Generations: Transformations in Europe, Edward Elgar, Aldershot,
pp. 210–218.
Yeandle, S. (2007) Carers, employers and the state: clarifying the social contract, Paper presented
to the Caregivers: Essential Partners in Care Conference (FICCDAT), Toronto, Canada, 16–19
June.
Yeandle, S., Bennett, C., Buckner, L., Suokas, A. and Shipton, L. (2006a) Who Cares Wins: the
Social and Business Benefits of Supporting Employed Carers, Carers UK, London.
Yeandle, S., Buckner, L. and Shipton, S. (2006b) Local Challenges in Meeting Demand for Domi-
ciliary Care – Synthesis Report, Centre for Social Inclusion, Sheffield Hallam University,
Sheffield.
Zimmermann, A.C. and Easterlin, R.A. (2006) Happily ever after? Cohabitation, marriage, divorce,
and happiness in Germany, Population and Development Review, 32(3): 511.
Chapter 2
Delayed Childbearing and Childlessness
Roona Simpson
Introduction
Fertility patterns have changed significantly in most European countries since the
1960s. Although there are important differences both between and within countries
in the pace of decline, by the 1990s fertility reached below replacement levels across
Europe, with several countries experiencing lowest-low levels (less than 1.31 chil-
dren per woman) (European Commission 2006a). Declining fertility is one aspect
of an inter-related package of demographic trends that both reflect and constitute
a transformation of the context in which people may or may not partner and par-
ent. Debates within demography over the causes and consequences of the second
demographic transition have been paralleled by sociological debate on the impact
of individualism on personal life. This chapter considers these debates in relation to
findings from analysis of the British cohort studies looking at changes over time in
the propensity to delayed childbearing.
Within demographic debate a ‘post-materialist values’ perspective contends that
changes in social and demographic behaviour have been driven by a growth in
values of self-realisation in the context of a decline in tradition and diminishing con-
straints (Van de Kaa, 1987; Lesthaeghe, 1995; Leete, 1998). Similarly, much soci-
ological theorising on the changing character of intimacy and partnership empha-
sises an increase in behavioural options in the context of increasing individualism
(Bauman, 2000; Giddens, 1992; Beck and Beck-Gernsheim, 1995, 2002). Within
these debates, increasing childlessness has featured as a signifier of a ‘choice biogra-
phy’ or, in more pessimistic accounts, as indicative of self-interested individualism.
Whereas the decline of fertility below replacement level was previously viewed
as the most important feature of the demographic transition (Van de Kaa, 1987),
the postponement of first births is now portrayed as the most radical transformation
(Lesthaeghe and Neels 2002: 33). While second and higher-order births are being
postponed as well, this is mostly as a consequence of first-birth postponement rather
R. Simpson (B)
Centre for Research on Families and Relationships, University of Edinburgh, UK
e-mail: Roona.Simpson@ed.ac.uk

24 R. Simpson
than a manifestation of change in birth intervals (Sobotka, 2004). There is some

debate in the literature as to the extent to which delayed childbearing is related to
increased childlessness, or constitutes a ‘postponement transition’ towards a late
fertility regime (Kohler et al., 2002: 659–661).
There are numerous factors proposed in the literature to explain delayed child-
bearing and much empirical research investigating their impact. These include the
spread of access to modern contraception (e.g. Murphy, 1992, 1993), changes in
participation in education (e.g. Kohler et al., 2002) and employment (e.g. Brewster
and Rindfuss, 2000), the wider economic context (e.g. Bagavos and Martin, 2000)
as well as the shifting character of intimacy and partnership relations in the con-
text of shifts in balances of power between men and women (e.g. Lesthaeghe, 1995;
Oppenheimer, 1994). However, many factors are clearly interconnected, with their
relative impact difficult to quantify. Thus, in relation to contraception there are
varying emphases on the interpretation of this as a technical factor addressing a
demand for birth control (e.g. Castles, 2003) or as a precursor to broader behavioural
and cultural changes, including changing partnership patterns more generally (e.g.
Van de Kaa, 1994: 114). As well as understanding this as a necessary precondition
for various other changes, some authors emphasise the social and psychological sig-
nificance for women of the ability to control their fertility (Presser, 2001; Hakim,
2003). The expansion of post-secondary education in recent decades in a context
of an increased emphasis on educational attainment in securing stable employ-
ment, adequate income and career development, has been interpreted as impacting
directly on fertility. ‘Being in education’, during which students usually lack ade-
quate resources and with future employment conditions uncertain, strongly reduces
the likelihood of having a first child (e.g. Blossfeld, 1995; Hoem, 2000). However,
as well as delay attributable to time spent in education, this influences the timing of
parenthood indirectly, through increased career opportunities or a less traditional or
family-centred value orientation (Oppenheimer, 1994).
Educational level is clearly linked to employment, associated with an enhanced
position in the labour market. Much attention has focused on the demise of a ‘male
breadwinner model’ and the increasing likelihood of women combining caring for
children with participation in the labour market. Debates about the opportunity costs
of childbearing for women and conflict between career aspirations and motherhood
are prevalent in the demographic literature. The ‘positive turn’ in many countries,
with higher fertility associated with higher female labour force participation, has
drawn attention to specific policy contexts within which individuals seek to balance
work and family aspirations (Esping-Anderson, 1999; McDonald, 2000a; Castles,
2003; Hobson and Olah, 2006a). Birth timing and spacing may be key strategies
adopted to reconcile these (De Cooman et. al., 1987; Brewster and Rindfuss, 2000).
Empirical studies demonstrate substantial differences in first birth timing according
to the level of education, with highly educated women postponing childbearing to a
larger extent than women with less education who often continue to have children
at early ages (Gustaffson et al., 2002; Joshi, 2002; Sobotka, 2004).
There has been much recent sociological attention to new social risks conse-
quent on various inter-related changes occurring as part of a shift to a post-industrial
2 Delayed Childbearing and Childlessness 25
society. One aspect of this is the extent to which risk aversion applies to the personal
sphere and influences the likelihood of people embarking on partnership or parent-
ing (e.g Beck, 1999; Taylor-Gooby, 2004; Lewis, 2006). There has also been atten-
tion to the impact of global restructuring and welfare state retrenchment of recent
decades in demographic debates, with unemployment for example analysed as a fac-
tor that increases economic uncertainty and discourages union formation and parent-
hood (see Kohler et al., 2002; Meron and Widmer, 2002; Hobson and Olah, 2006b).
Sobotka argues that the available evidence suggests the influences of uncertainty on
first birth timings differ in time, across countries, by types of uncertainty and with a
differing impact on various population groups (Sobotka, 2004: 21), citing in illustra-
tion a study of fertility in England and Wales (De Cooman et al., 1987) which found
fertility reactions to economic change varying by stage in family formation, with
labour market conditions influencing timing rather than the quantum of fertility.
Various inter-related changes have resulted in considerable transformations in
family and household formations, as a now familiar list of trends including the
delay and decline in marriage, the separation of marriage and reproduction (both
childless marriage and non-marital fertility) and an increase in unconventional liv-
ing arrangements such as solo living, demonstrate. These are occurring alongside
a relative lack of change in some traditional elements of partnership and parent-
ing, including in gendered divisions of labour (see Folbre, 1994; Hoshchild, 2001).
Trends to delayed childbearing and increased childlessness in recent decades are
clearly not independent of trends in partnership formation and dissolution. There is
empirical data supporting the idea that more complex partnership patterns are asso-
ciated with late entry into parenthood (e.g. Oláh, 2005). It is, however, transforma-
tions in ways of interacting within relationships, rather than changing structures per
se, which are argued as having profound consequences for fertility. Whereas some
authors contend these changes reflect an increasing emphasis on self-fulfilment
(e.g. Van de Kaa, 2004), others have interpreted this in terms of women’s greater
independence allowing them to set higher standards in relation to partnership (e.g.
Oppenheimer, 1994), including expectations around gender equity in the domestic
sphere (e.g. McDonald, 2000b; Berrington, 2004; Wasoff and Dey, 2007).
There is considerable debate around which are the most significant factors driv-
ing contemporary changes in both partnership and parenting, and how it may be
possible to identify primary factors (Kohler et al., 2002; Caldwell and Schindlmayr,
2003, Hobcraft, 2004). Several authors have also questioned individualist expla-
nations of demographic behaviour as the outcome of a ‘freeing’ of agency from
changing social conditions, arguing rather that individual choices are inseparable
from the social contexts, including the structural process and cultural mores, within
which they are made (Irwin, 2005; Smart, 2007).
Most research on fertility focuses on women, to the empirical neglect of men
(Greene and Biddlecom, 2000: 81). This may be in part due to the greater reliabil-
ity of reported fertility by women. Previous research suggests some unreliability in
men’s reporting of children, especially of non-marital births; for example Rendall
et al. (1999). An evaluation of men’s retrospective fertility histories from the British
Household Panel Study found only 60 of men’s births from a previous marriage
26 R. Simpson
were reported per 100 of women’s births, with women’s fertility reports matching
birth registration statistics (Rendall et al., 1999). Survey data do not lend them-
selves readily to explorations of the motivations and meanings individuals ascribe
to statuses such as childlessness. Yet qualitative research studies on the experience
of childlessness in recent decades similarly focuses on women (see Veevers, 1980;
Cameron, 1997; Lisle, 1999). Such a focus, as well as neglecting the experience of
men, can reinforce a view of women as primarily reproductive (see Watkins, 1993).
Furthermore, despite an emphasis in much of the sociological and demographic lit-
erature on ‘choice’ biographies, the emphasis on women risks ignoring the extent
to which preferences and decision-making around parenthood are negotiated at the
level of the couple. This reinforces the importance of considering contemporary
change in the partnership context in which decisions about childbearing are being
made.
In common with many other industrialized nations, there have been significant
changes in the timing and context of partnership and parenthood in Britain in recent
decades. In 1971, the average age of first marriage in England and Wales was 25
for men and 23 for women. However, by 2003, this had increased to 31 for men
and 29 for women. Alongside this delay in marriage there has been a huge increase
in cohabitation. Non-marital cohabitation amongst those under 60 in Great Britain
doubled between 1986 (the earliest year for which data are available on a consis-
tent basis) and 2006, from 11% to 24% for men, and 13 to 25% for women (ONS,
2008). There has also been a dramatic increase in recent decades of men and women
living alone at ages conventionally associated with co-resident partnership and par-
enthood: over the past two decades the proportion of people living alone has dou-
bled, from 7% to 14% of men aged 25–44 and from 4% to 8% of women in this age
group (ONS, 2008: 19). These all have implications for the context of childbearing.
Births by unmarried women accounted for 42% of all births in the United Kingdom
in 2004, compared to an EU (15 countries) average of 33% (European Commission,
2006b). For women, the average age of first birth rose from 24 in the early 1970s to
28 in 2006, and this is even higher for first births within marriage, with an average
age of 30 in 2006 (ONS, 2008: 24). There has also been an increase in delayed child-
bearing and childlessness. Whereas around 16% of women born in 1931 remained
childless at 35, this compares with a quarter of women born in 1971. The proportion
of women remaining childless at 45 has risen from 14% of women born in 1931 to
just under a fifth (19%) of those born in 1961, the latest cohort to reach the end of
their childbearing years (ONS, 2008: 24).
Figure 2.1 illustrates changes over time in the percentages of women remaining
childless, having only one child, or large families (four or more children). Figures
for 1920 are included to demonstrate that these figures are not unprecedented, with
an average age of first birth of 28.7 for women born in 1920.
The previous discussion outlines the predominant factors identified in the liter-
ature with delayed childbearing. However, as noted above, there is relatively little
attention to that of men. The following section reports analysis of the British cohort
studies looking at men and women born in 1958 and 1970. As well as comparisons
by sex, this inter-cohort analysis allows comparisons across time, thereby providing
Fig. 2.1 Completed family 25

No children 1 child 4 or more children
size, selected cohorts (%)
(Source: ONS Cohort Series 20
Woman (%)
FM1 No. 35, Table 10.5) 15
10
0
1920 1940 1945 1950 1955 1960
Year
a picture of the impact of social change on the lives of cohort members. Previous
analyses of the cohort studies demonstrate changes across a range of domains such
as education, employment, health and family life, which reflect the wider social
context in which these changes were experienced (Ferri et al., 2003). Those born
in 1958 will have experienced their early adulthood during the 1980s, a very dif-
ferent political, social and economic context from those born in 1970: whereas the
1958 cohort initially experienced a buoyant labour market but faced worse condi-
tions as they got older, the opposite occurred for those born in 1970. Rising living
standards in recent decades have been accompanied by increasing polarisation in
terms of wage inequality. Cohort members remaining childless in their early thirties
will also have experienced this in differing circumstances, those born in 1958 in the
early 1990s, and those born in 1970 in the early twenty-first century.
Data and Methods

The National Child Development Study (NCDS) and British Cohort Study (BCS)
are longitudinal surveys following a sample of individuals born in Great Britain
between 3–9 March 1958 and 5–11 April 1970 respectively. Surveys to monitor the
educational, physical and social development of these cohorts have taken place at
varying intervals, with data collected at birth, 7, 11, 16, 23, 33, 41/42 and 46/47for
the NCDS and at birth, 5, 10, 16, 26, 29/30 and 33/34 for the BCS. Earlier stages
of the studies were conducted by various organisations, however data collection
for these and other birth cohort studies is now the responsibility of the Centre for
Longitudinal Studies, Institute of Education, University of London. Full fertility
histories were produced for the NCDS cohort at age 33/34 (see Di Salvo, 2005), and
these were drawn on in the analysis presented below.1
Longitudinal data enabling the opportunity to investigate demographic events
within a life course perspective, however, are subject to attrition, and approximately
1 Fertilityhistories are also collected at age 23 years for the NCDS cohort; although a cross-
sectional approach was favoured for the analysis of NCDS data in this study.
28 R. Simpson
two thirds of those taking part in the birth survey were interviewed in their early thir-
ties. Previous analyses of the response bias arising from this suggests it is the most
socio-economically disadvantaged that are over-represented amongst those lost over
time (for further detail see Shepherd, 1995; Plewis et al., 2004). Both samples also
under-represent those who began child-bearing in their teens (Berrington, 2003).
The analysis assumes any attrition bias is similar in both surveys.
There are 11,407 men and women aged 33 in the NCDS5 cohort, and 11,261 aged
30 in the BCS 2000 cohort. The findings presented below report analysis conducted
on individuals in these sweeps who were also present at ages 33/34 (BCS 2004) (n =
9,665), in order to conduct comparisons by cohort at exact age 33. Survival analyses
were conducted on both cohorts, subsequent to cleaning the data to address those
with missing/incomplete birth dates. As the focus of this chapter is on postponed
fertility it only considers biological children, and childlessness is defined as not
having had a live birth; those classified as childless may include respondents with
adopted or step-children.
The analysis considered several variables identified as important in the litera-
ture.2 As noted above, previous research has attributed fertility postponement to
changes in education and employment (particularly for women), and partnership
formation. However, these are mediated by social class (Berrington, 2003). The fol-
lowing section reports the results of analyses looking at childlessness by sex in
relation to social class background identified using the occupational social class
(Registrar General’s definition) of the cohort member’s father (or father figure) at
age 16; educational attainment by age 33– those with no qualifications, school-
level qualifications (including A/S level and their NVQ equivalents) and tertiary
level qualifications, including Higher qualifications, degrees and NVQ equivalents
(4–6); current economic activity-looking at those working full-time or part-time
(whether employee/self-employed), unemployed and seeking work, permanently
sick/disabled, or looking after home/family; and current marital and (co-residential)
partnership status.3
2 Initialanalysis was carried out looking at religiosity. The Second Demographic Transition
attributes declining fertility to, inter alia, a growth of values of self-realisation and freedom from
traditional forces of authority such as religion and previous research reports a correlation between
increased religious activity and more traditional attitudes to family (e.g. Berrington and Diamond,
2000; Oláh, 2005). However, while there was a decline in religiosity evident (only 11 per cent
of the 1970-born cohort who defined themselves as having a religion attended services at least
monthly, compared with 30 per cent of the 1958 cohort), there was no significant association with
remaining childless in one’s early thirties.
3 Marital status distinguishes the currently married (whether first or subsequent), the previously
married (the separated or divorced, as well as the very small numbers widowed in each cohort),
and the never-married. Partnership status distinguishes those currently cohabiting: partnership his-
tories record relationships in which respondents ‘lived with someone as a couple’ for a month or
more. However, as Murphy (2000) observes, assuming all co-residential partnerships as equivalent
raises several issues: these may differ not just in de jure status, but in terms of the meanings and
motivations these hold for respondents.
The Propensity to Childlessness Among Young

Adults – Differences by Sex and Changes Over Time
The analysis compares the propensity amongst men and women to remain child-
less by exact age 33, considering this in relation to various socio-economic charac-
teristics and circumstances. Comparisons are also made with those who have had
children at the same age, to provide a broad overview. As noted above cohort mem-
bers who remain childless at 33 will have experienced this in somewhat different
contexts.
Delayed Childbearing by Sex
As Figure 2.2 illustrates, there are significant differences by sex, with larger propor-
tions of men remaining childless at age 33 compared with women in both cohorts.
Thus, a third of men in the 1958 cohort remained childless at 33 compared to just
under a quarter (24%) of women at the same age. For the 1970 cohort the figures
were over half (51%) of men compared with less than two fifths (36%) of women at
the same age.
These figures also illustrate a dramatic increase in childlessness amongst adults
in their early thirties in just 12 years, particularly for men. The median age of first
birth for women has increased from 27 to 30 and for men from 30 to 34. Previous
analysis of the cohort studies (Berrington, 2003) finds that, alongside an increase in
median age across cohorts, similar proportions (around 10%) became teenage moth-
ers, indicating increasing polarisation between women in relation to age at first birth.
Delayed Childbearing by Social Class Background

As Figure 2.3 illustrates (see also Table 2.1), there are significant differences in
the propensity to remain childless at age 33 by social class background, defined
60
Male Female
50
Childless (%)
40
30
51
20 36
33
Fig. 2.2 Childlessness at 33 10 24
by sex and cohort (%)
(Source: Author’s 0
1958 1970
calculations based on survey
data) Year
30 R. Simpson
80
I and II IIInm IIIm IV and V No father figure
70
60
Childless (%)
50
40
30
20
10
0
Male Female Male Female
1958 1970
Fig. 2.3 Childlessness at 33 by social class background (%)

(Source: Author’s calculations based on survey data)
by occupational social class of the cohort member’s father (or father figure) at the
time of birth. Considering firstly those born in 1958, the differences are largest
for women, with a third from professional or managerial backgrounds (I and II)
remaining childless at 33 compared to less than a fifth (18%) of those with fathers
Table 2.1 Men and women remaining childless at 33 by cohort
1958 1970
Male % Female % Male % Female %
Social class background (father’s occupation at age 16)

I and II 38 33 58 43
III non-manual 37 28 54 39
III manual 33 22 48 33
IV and V 29 18 44 30
No father figure 38 21 43 32
Highest educational attainment (by age 33)
No qualifications 29 12 45 25
School level qualifications 31 21 46 29
Tertiary qualifications 40 37 61 48
Current partnership status
Married 17 13 33 23
Previously married 35 21 41 29
Cohabiting 50 40 55 47
Single never married 96 80 89 67
ALL 33 24 51 36
Total (100%) 5,577 5,791 4,237 4,765
Source: Author’s calculations based on survey data.

in semi-skilled and unskilled occupations (IV and V). The differences for men
of the same age across social class background are smaller, with the same pro-
portion of men (38%) with no father figure at birth as those from professional
or managerial backgrounds remaining childless at 33, with men with fathers in
semi-skilled and unskilled occupations (IV and V) those least likely to remain
childless (29%).
Looking at the 1970 cohort, here the differences are more extreme amongst men,
a difference of 15 percentage points: 58% from professional or managerial back-
grounds remaining childless at 33 compared with 43% of men with no father figure –
those least likely to be childless from this cohort. The pattern for women born in
1970 remains the same as for those from the 1958 cohort, with higher proportions
from professional or managerial backgrounds remaining childless at over-two fifths
(43%), and those with fathers in semi-skilled and unskilled occupations least likely
to remain childless (30%) at this age. The changes over time however are largest
for men from professional or managerial backgrounds, an increase of 20 percentage
points between cohorts.
Childlessness by Educational Attainment
Recent decades have seen a massive expansion in post-secondary education, with

young adults spending an increasing proportion of time in education. The 1958
cohort experienced an educational context in which the majority (over 60%) of
young people left school at 16, however by the time the 1970 cohort reached 16,
in 1986, this had declined to around half (54% of males and 45% of females)
(Makepeace et al., 2003: 41). Analysis comparing highest educational achievement
by sex across the cohorts, not illustrated here, shows higher proportions of both men
and women gaining tertiary qualifications by their early thirties, over a third of the
1970s cohort compared with a quarter for those born in 1958. As mentioned above
there is a considerable literature addressing the influence of educational attainment
on fertility, with a large body of empirical evidence of substantial differences in first
birth timing according to levels of education (Sobotka, 2004). There has been much
attention to increases in women’s educational attainment in particular. Whereas for
the 1958 cohort slightly higher proportions of men than women had tertiary edu-
cational qualifications by age 33 (28 versus 25%), amongst the 1970 cohort this
was reversed, with 35% compared with 32% of men. However, as Makepeace et al.
(2003) observe, this broad observation does not address continuities in the large
differences by gender in subjects studied. Furthermore, their analysis of the cohort
studies demonstrate that the gap in the chances of gaining tertiary qualifications for
those from the highest and lowest social classes has widened steadily over time.
Nevertheless, differentials in delayed childbearing by educational level are also
evident for men. Table 2.1 presents figures for the proportion of men and women
remaining childless in their early thirties, and differentials by educational attain-
ment are significant for both men and women (Fig. 2.4).
32 R. Simpson
80
No qualifications School level qualifications Tertiary qualifications
70
60
Childless (%)
50
40
30
20
10
0
1958 1970
Fig. 2.4 Childlessness at 33 by educational qualifications (%)

Looking at the 1958 cohort, nearly two fifths (37%) of women with tertiary qual-
ifications remain childless at 33, compared with only 12% of women of the same
age with no qualifications. While the differences are less stark between men, what is
noteworthy is the difference by sex, with unqualified men more than twice as likely
to be childless at this age compared with unqualified women. This differential by
sex is also evident for the 1970 cohort: 25% of women with no qualifications remain
childless at 33, compared with 45% of unqualified men. This difference by sex is
also marked amongst those with tertiary education; where just under half (48%) of
women in this cohort remain childless at 33, compared with nearly two thirds of
men (61%).
Childlessness by Current Economic Activity
There has been much debate in the literature over the role of women’s employment
in particular in fertility decline. Several commentators propose women’s increased
labour force participation as an indicator of increasing gender equality; such claims
however do not address continuing differences in hours of work or levels of pay
(Perrons, 2002, 2003). Previous research by Makepeace et al. (2003) finds differ-
ences by sex in the proportions of men and women born in 1970 who experienced
continuous employment between ages 16 to 30 to be 30% compared with 14%
respectively. There is also an increasingly strong relationship between qualifications
and continuous employment: in the 1970 cohort, highly qualified women are more
likely to remain in continuous employment than both their less qualified contempo-
raries and their predecessors in the 1958 cohort (Makepeace et al., 2003: 58).
Preliminary analysis of the 1958 cohort’s economic activity in their early thir-
ties showed significant differences in economic status by sex: 89% of all men were
working full-time, compared with 36% of women; another third of women worked
part-time (32%) while others looked after the home and family full time (28%).
Similar analysis on the 1970 cohort in their early thirties shows an increase in
the proportion of all women in full-time employment, to 43%, with a concomitant
decrease in the proportions looking after home and family (19%). Similar propor-
tions as in the previous cohort were working part-time (32%). However, for men,
there is little change in the proportions working full time, around 90% in both
cohorts.
Subsequent analyses considered the incidence of childlessness in relation to eco-
nomic activity by parental status. Not surprisingly, there were significant differences
evident amongst women, for both cohorts. Amongst the 1958 cohort, the vast major-
ity who remained childless in their early thirties were in full-time employment (83%
compared to 22% of mothers). The picture for men was very different however, with
a slightly higher proportion of fathers (91% compared with 87% childless) working
full-time. Looking at the 1970 cohort indicates very little change over time for men,
however there are changes over time for women. The proportions of women who
remained childless in their early thirties and who were working full-time declined
compared with the earlier cohort, now standing at 77%. The proportions of childless
women working part-time increased slightly, from 7% amongst the 1958 cohort to
11% of the 1970 cohort. Amongst mothers of the same age, there have also been
changes over time, with slightly higher proportions working full-time (25% among
the 1970 cohort) and part-time (from 39% amongst the 1958 cohort to 43%) of the
1970 cohort, with a concomitant decrease in the proportions categorised as looking
after home and family full time, from over a third (35%) of those born in 1958 to a
quarter (26%) of those born in 1970 (see also Hansen et al., this volume).
British men have the highest average number of working hours compared with
other men in the European Union, while British fathers work longer hours on aver-
age than men without dependent children (Kiernan, 1998). This analysis shows
men remain overwhelmingly in full-time employment, regardless of parental status.
While the proportions of working mothers has increased, these figures comparing
changes over time in economic activity suggest a relative lack of change in respon-
sibilities for childcare between men and women.
Childlessness by Partnership Status

Previous analyses of the cohort studies have highlighted the considerable decline in
marriage in just twelve years: whereas three quarters of women and two thirds of
men had married by age 29, just over half of women and a little over a third of men
born in 1970 had done so (Berrington, 2003). Some of the decline can be explained
by an increase in cohabitation, and cohabitation was the most common form of first
partnership for the 1970 cohort (Berrington, 2003). However, there is also evidence
of a delay in partnership, whether marriage or cohabiting, alongside an increase in
relationship dissolution. Women born in 1970 were much less likely to have entered
their first partnership at a very young age (just over a quarter had lived with a partner
34 R. Simpson
by age 20 compared with 40% of women born in 1958), while twice as many men
and women born in 1970 have had at least one previous relationship by age 30; of
those born in 1948, the majority (about 8 out of 10) were living with a partner by
age 42 (Smith and Ferri, 2003). Nevertheless, these changes indicate a concomitant
increase in an experience of singleness, either prior to or between relationships.
Changes over time in living arrangements also indicate a greater decline in res-
idential partnership amongst men than women, with a higher proportion of men
either living alone (14% compared to 9% of women born in 1970) or living in the
parental home (one in six men born in 1970, more than twice as many compared to
women of the same age) (Ferri and Smith, 2003). Analyses of the British Household
Panel Survey (BHPS) note an increasing tendency since the 1980s for young adults
to return to the parental home (Ermisch, 1996).
In part this may be related to changes such as the increase in partnership disso-
lution. However, this may also be associated with wider changes in areas such as
housing or education, and indicates a delay in young adults establishing themselves
as economically independent. Analyses of the cohort studies identify a significant
decline in home ownership over time. Slightly higher proportions of women born
in 1970 compared to men were buying with a mortgage in their early thirties (62
versus 58%), a reversal of the situation for the 1958 cohort (75% women compared
with 78% men). While the proportion living in social housing (housing associa-
tions or local authority) remained the same at about 15%, the proportions of those
privately renting had increased over time (from 6% to 21%) for the 1970 cohort,
while the proportions classified as ‘living rent free’, including with parents, had
also increased.
Analysis of the cohorts by legal marital status indicates a considerable increase
in the proportions remaining never-married over time, as well as significant differ-
ences by sex. Just over a fifth (22%) of all men from the 1958 cohort remained
never-married in their early thirties, compared to 15% of women of the same age.
However, this had increased to 40 and 31% of the 1970 cohort respectively. Differ-
ences between men and women may be explained in part by the fact that traditionally
women have tended to marry at younger ages than men. As this category does not
distinguish cohabiting, subsequent analysis considered current partnership status,
distinguishing those currently cohabiting regardless of previous marital status.
Comparisons of the cohorts in their early 30s indicate considerable change in
partnership patterns, including the decline in marriage and increase in cohabiting.
Whereas four-fifths of the women born in 1958 were or had been married (72%
currently married with a further 8% previously married), this compared with less
than two-thirds of the 1970 cohort (57 and 7% respectively). Alongside an increase
in cohabitation (from around 10% of men and women in the 1958 cohort to around
20% of those born in 1970), there was also an increase in the proportions categorised
as currently single and never-married, from 15% to 22% of men and 10% to 17% of
women.
Table 2.1 shows figures for the proportions of men and women who remain child-
less at age 33 by current partnership status. The proportion of men from the 1958
cohort who have not had children at 33 ranges from less than a fifth (17%) of married
Male
1
Childless proportion
0.8
0.6
0.4
0.2
0
16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
Age
Married Previously married Cohabiting Single never married
Female
1
0.8
0.6
0.4
0.2
0
16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
Age
Fig. 2.5 Childless men and women by partnership status, 1958 cohort
men to 96% of those never-married and currently single. Figures for childlessness
amongst women of the same cohort range from 13% of married women to four-fifths
of the single never-married.
Figure 2.5 illustrates these changes, showing the relative decline in childlessness
as cohort members age. The graphs show that it is the previously-married amongst
both sexes who have the highest likelihood of having children at earlier ages. How-
ever, by age 33, the lowest proportions of childlessness are amongst the currently
married. For both sexes it is those who are single never-married who are most likely
to remain childless at this age.
Figures for the 1970 cohort show that while childlessness at age 33 has increased
generally, the broad pattern persists (Table 2.1). Given the changes in partnership
patterns over time, we might expect the significance of marital status to wane. Yet,
despite changes such as an increase in cohabitation, there is a persistent association
between marriage and childbearing.
36 R. Simpson
Male
1
0.8
0.6
0.4
0.2
0
16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
Age
Married Previously married Cohabiting Single never married
Female
1
0.8
0.6
0.4
0.2
0
16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
Age
Fig. 2.6 Childless men and women by partnership status, 1970 cohort
Nevertheless, the proportions of both men and women who are currently married
and childless have nearly doubled over time, from 17% to 33% of men and from
13% to 23% of women. The incidence of remaining childless at age 33 amongst
other categories has increased far less, and has decreased amongst those who are
single never-married at 33, from 96% to 89% of men and from 80% to 67% of
women.
The graphs in Fig. 2.6 illustrate the delay in childbearing relative to the earlier
cohort, with the proportions remaining childless at specific ages starting to decline
mostly from the mid-twenties onward for both sexes. There is then a more rapid
decline in the late twenties.
Thus, while the later cohort indicates that there is a narrowing of the differentials
by partnership status over time, these differences persist. This suggests that the pace
of shifts in the context of childbearing (within marriage) lag behind shifts in part-
nership formation, a finding which has implications for future fertility patterns. The
decline in marriage may herald a shift to cohabitation, while the delay in partnership
may reflect postponement rather than overall decline. However, there is considerable
evidence that cohabitation is a more fragile relationship than marriage, and that
divorce is more common in marriages preceded by cohabitation (e.g. Haskey, 1992,
Ermisch and Francesconi, 1996). Previous research on the 1958 cohort (Berrington
and Diamond, 2000) suggests that experience of independent living prior to part-
nership formation is associated with a preference for cohabitation rather than direct
marriage, and higher separation rates amongst cohabiters. Relationship stability is
related to both timing and type of relationship. This suggests that current patterns of
fertility decline may be unlikely to alter dramatically in the future, given trends in
partnership formation and dissolution and the association of childbearing and mar-
riage identified above. The figures on partnership patterns also indicate the impor-
tance of further research on the gendered nature of these trends; while there has
been a recent increase in research on singleness amongst women for example, there
is little contemporary research on singleness amongst men (Simpson, 2005).
Conclusions
The analysis presented in this chapter seeks to identify changes over time in the
timing and propensity to childbearing in Britain in relation to various factors iden-
tified as important in the literature. There is much attention in the sociological and
demographic literature to notions of increasing individualisation and values of per-
sonal autonomy; nevertheless, in demonstrating consistent differences by factors
such as social class background and educational attainment across cohorts, this anal-
ysis underlines the continuing influence of social structural processes.
However, these differences provide a descriptive context that highlights the need
for more attention to the ways in which changes in demographic behaviour at the
level of the individual or the couple are related to wider social processes. Authors
such as Irwin argue for a more nuanced understanding of structure, not just in terms
of choice and constraint, but in terms of changing contexts within which people
make choices, and in which some courses of action are much more likely to be
‘chosen’ than others (Irwin, 2005: 179). Such scholarship emphasises the impor-
tance of considering the iterative relationship between individual motivations and
behaviours and the particular social and historic context in which these are shaped.
In looking at both women and men, the analysis presented here demonstrates
similarities not evident in research looking only at women, for example the associ-
ation of tertiary educational qualifications with delayed childbearing amongst both
women and men. It also however illustrated the differential impact of gender, for
example on working practices, irrespective of parenthood status. There is increasing
attention to the need to regard gender as an aspect of identity that is relational, rather
than as a fixed attribute of individuals (see Watkins, 1993; Greene and Biddlecom,
2000). As well as attention to differences between socio-economic contexts and cul-
tural mores, there is a need to look further at differences between men and women
38 R. Simpson
in terms of both subjectivities as well as practices, in order to understand fertility

behaviour more fully.
References
Bagavos, C. and Martin, C. (2000) Low fertility, families, and public policies, Synthesis Report
of the Annual Seminar, The European Observatory on the Social Situation, Demography and
Family, Seville, 15–16 September.
Bauman, Z. (2000) Liquid Modernity, Polity Press, Cambridge.
Beck, U. (1999) World Risk Society, Polity Press, Malden, Mass.
Beck, U. and Beck-Gernsheim, E. (1995) The Normal Chaos of Love, Polity Press, Oxford.
Beck, U. and Beck-Gernsheim, E. (2001) Individualization: Institutionalized Individualism and its
Social and Political Consequences, Sage, London.
Berrington, A. (2003) Change and continuity in family formation among young adults in Britain,
SSRC Working Paper A03/04, University of Southampton, Southampton.
Berrington, A. (2004) Perpetual postponers? Women, men’s and couple’s fertility intentions and
subsequent fertility behaviour, Population Trends, 117, Autumn.
Berrington, A. and Diamond, I. (2000) Marriage or cohabitation: a competing risks analysis of first-
partnership formation amongst the 1958 British birth cohort, Journal of the Royal Statistical
Society, Series A, 163: 127–151.
Blossfeld, H.-P. (1995) Changes in the process of family formation and women’s growing eco-
nomic independence: A comparison of nine countries, In Blossfeld, H.-P. (ed.) The New Role
of Women. Family Formation in Modern Societies. Westview Press, Boulder, pp. 3–32.
Brewster, K.L. and Rindfuss, R.R. (2000) Fertility and women’s employment in industrialized
nations, Annual Review of Sociology, 26, 271–96.
Caldwell, J. and Schindlmayr, T. (2003) Explanations of fertility crisis in modern societies: a search
for commonalities, Population Studies, 57(3): 241–263.
Cameron, J. (1997) Without Issue: New Zealanders Who Choose to not have Children, Canterbury
University Press, Christchurch NZ.
Castles, F. G. (2003) The world turned upside down: below replacement fertility, changing pref-
erences and family-friendly public policy in 21 OECD countries, Journal of European Social
Policy, 13: 209–227.
De Cooman, E, Ermisch, J. and Joshi, H. (1987) The next birth and the labour market: a dynamic
model of births in England and Wales, Population Studies, 41(2): 237–268.
Di Salvo, P. (2005) Child histories – reconciling self-completion and interview data, NCDS5 Data
Note 10, Centre for Longitudinal Studies, London, http://www.cls.ioe.ac.uk.
Ermisch, J. (1996) Analysis of leaving the parental home and returning to it using BHPS data,
Working Paper 96–1, Working Papers of the ESRC Research Centre Micro-Social Change,
University of Essex, Colchester.
Esping-Anderson, G. (1999) Social Foundations of Postindustrial Economies, Oxford University
Press, Oxford.
European Commission (2006a) Eurostat Indicators, Total Fertility Rates Eurostat, Statisti-
cal Office of the European Communities Unit F1: Demographic and migration statistics,
http://epp.eurostat.ec.europa.eu
European Commission (2006b) Eurostat Indicators, Live Births outside Marriage. Eurostat, Sta-
tistical Office of the European Communities Unit F1: Demographic and migration statistics,
http://epp.eurostat.ec.europa.eu
Ferri, E. Bynner, J. and Wadsworth, M. (eds.) (2003) Changing Britain, Changing Lives: Three
Generations At The Turn Of The Century, London: Insitute of Education, University of London,
London.
Ferri, E. and Smith, K. (2003) Partnerships and parenthood, in Ferri, E, Bynner, J. and Wadsworth,
M. (eds.) Changing Britain, Changing Lives: Three Generations at the Turn of the Century,
Institute of Education, University of London, London.
Folbre, N. (1994) Who Pays for the Kids? Gender and the Structures of Constraint, Routledge,
London.
Giddens, A. (1992). The Transformation Of Intimacy. Sexuality, Love and Eroticism in Modern
Societies, Polity Press, Cambridge.
Greene, M. and Biddlecom, A. (2000) Absent and problematic men: demographic accounts of male
reproductive roles, Population and Development Review, 26(1): 81–115.
Gustaffson, S.S, Kenjoh, E. and Wetzels, C. (2002) The role of education on postponement of
maternity in Britain, Germany, the Netherlands and Sweden, in Ruspini, E.and Dale, A. (eds.)
The Gender Dimension of Social Change, The Policy Press, Bristol.
Hakim, C. (2003) A new approach to explaining fertility patterns: preference theory’, Population
and Development Review, 29(3): 349–374.
Haskey, J. (1992) Pre-marital cohabitation and the probability of subsequent divorce: analyses
using new data from the General Household Survey, Population Trends, 68: 10–9.
Hobcraft, J. (2004) Method, theory and substance in understanding choices about becomgin a par-
ent: progress or regress? Comment on Caldwell and Schindlmayr’s ‘Explanations of the fertility
crisis in modern societies: a search for commonalities’, Population Studies, 58(1): 81–84.
Hobson, B. and Olah, L.Sz. (2006a) Birthstrikes? Agency and capabilities in the reconcilia-
tion of employment and family, Marriage and Family Review, No. 39, The Haworth Press,
Binghamton NY.
Hobson, B. and Olah, L.Sz. (2006b) The positive turn or birthstrikes? Sites of resistance to residual
male breadwinner societies and to welfare state restructuring, Recherche et Previsions No. 83
(March), Special Issue: Kind and Welfare State. Reforms of Family Policies in Europe and
North America, CNAF.
Hoem, B. (2000) Entry into motherhood in Sweden: the influence of economic factors on
the rise and fall in fertility, 1986–1997, Demographic Research, online journal available at
http://www.demographic-research.org/Volumes/Vol2/4.
Hoshchild, A. (2001) The Time Bind: When Work Becomes Home and Home Becomes Work, Henry
Holt, New York.
Irwin, S. (2005) Reshaping Social Life, Routledge, London.
Joshi, H. (2002) Production, reproduction and education: women, children and work in contempo-
rary Britain, Population and Development Review, 28(3): 445–474.
Kiernan, K. (1998) Parenthood and family life in the United Kingdom, Review of Population and
Social Policy, 7: 63–81.
Kohler, H.-P., Billari, F.C and Ortega, J. (2002) The emergence of lowest-low fertility in Europe
during the 1990s, Population and Development Review, 28(4): 641–680.
Leete, R. (1998) Dynamics of Values in Fertility Change, Clarendon Press, Oxford.
Lesthaeghe, R. (1995) The second demographic transition in western countries: an interpretation,
in Mason, K. and Jensen, A.M. (eds.) Gender and Family Change in Industrialized Countries,
Clarendon Press, Oxford.
Lesthaeghe, R. and Neels, K. (2002) From the first to the second demographic transition: An inter-
pretation of the spatial continuity of demographic innovation in France, Belgium and Switzer-
land, European Journal of Population, 18(4): 325–360.
Lewis, J. (2006) Perceptions of risk in intimate relationships, Journal of Social Policy, 35(1):
39–57.
Lisle, L. (1999) Without Child, Routledge, New York.
Makepeace, G. Dolton, P., Woods, L., Joshi, H., and Galinda-Rueda, F. (2003) From school to the
labour market, in Ferri, E., Bynner, J. and Wadsworth, M. (eds.) Changing Britain, Changing
Lives, Institute of Education, London.
McDonald, P. (2000a) The “toolbox” of public policies to impact on fertility – a global view, Paper
presented at seminar on ‘Low Fertility, Families and Public Policies, organised by European
Observatory on Family Matters, Sevilla, Sept 15–16.
McDonald, P. (2000b) Gender equity in theories of fertility transition, Population and Development
Review, 26(3): 427–439.
40 R. Simpson
Meron, M. and Widmer, I. (2002) Unemployment leads women to postpone the birth of their first
child, Population-E 2002, 57(2): 301–330.
Murphy, M. (1992) Economic models of fertility in post-war Britain – a conceptual and statistical
re-interpretation, Population Studies, 46(2): 235–258.
Murphy, M. (1993) The contraceptive pill and women’s employment as factors of fertility change
in Britain 1963–1980: a challenge to conventional view, Population Studies, 47(2): 221–243.
Murphy, M. (2000) The evolution of cohabitation in Britain, 1960–1995, Population Studies,
54 (1): 43–56.
Oláh, L.Sz. (2005) First childbearing at higher ages in Sweden and Hungary: a gender approach,
Paper presented at the 15th Nordic Demographic Symposium, Aalborg University, Denmark,
28–30th April.
ONS (Office of National Statistics) (2008) Social Trends 38, Palgrave MacMillan, Hampshire.
Oppenheimer, V. K. (1994) Women’s rising employment and the future of the family in industri-
alised societies, Population and Development Review, 20(2): 293–342.
Perrons, D. (2003) The new economy and the work life balance, Gender Work and Organisation,
10(1): 65–93.
Perrons, D. (2002) Gendered divisions in the new economy: risks and opportunities, GeoJournal,
56: 271–280.
Plewis, I., Calderwood, L., Hawkes, D, and Nathan, G. (2004) Technical Report: Changes in the
NDCS and BCS7 Populations and Samples over Time, Centre for Longitudinal Studies, Institute
of Education, University of London, London.
Presser, H. (2001) Comment: a gender perspective for understanding low fertility in post-
transitional societies, Population and Development Review, 27: 177–183.
Rendall, M, Clarke, L, Peters, H. E, Ranjit, N, Verropoulou, G. (1999) Incomplete reporting of
male fertility in the United States and Britain: a research note, Demography, 36(1):135–144.
Shepherd, P. (1995) The National Child Development Study: an introduction to the origins of
the study and the methods of data collection, NCDS User Support Working Paper 1 (revised),
Social Statistics Research Unit, City University: London.
Simpson, R. (2005) Contemporary Spinsterhood in Britain: Gender, Partnership Status and Social
Change, PhD thesis, University of London, London.
Smart, C. (2007) Personal Life: New Directions in Sociological Thinking, Polity Press, Cambridge.
Smith, K. and Ferri, E. (2003) Housing, in Ferri, E., Bynner, J. and Wadsworth, M. (eds.) Changing
Britain, Changing Lives, Institute of Education, London.
Sobotka, T. (2004) Postponement of Childbearing and Low Fertility in Europe, PhD thesis, Uni-
versity of Groningen. Dutch University Press, Amsterdam.
Taylor-Gooby, P. (ed.) (2004) New Risks, New Welfare: The Transformation of the European Wel-
fare State, Oxford University Press, Oxford.
Van de Kaa, D.J. (1987) Europe’s second demographic transition, Population Bulletin, 42(1): 1–47.
Van de Kaa, D.J. (1994) The second demographic transition revisited: theories and expectations, in
Beets, G, van den Brekel, H, Cliquet, R, Dooghe, G, and de Jong Gierveld, J. (eds.) Population
and Family in the Low Countries 1993: Late Fertility and Other Current Issues, NIDI/CBGS
Publication, No. 30, Swets and Zeitlinger, Berwyn, Pennsylvania/Amsterdam, pp. 81–126.
Van de Kaa, D.J. (2004) The true commonality: in reflexive societies fertility is a derivative, Pop-
ulation Studies, 58 (1): 77–80.
Veevers, J.E. (1980) Childless by Choice, Butterworths, Toronto.
Wasoff, F. and Dey, I. (2007) Fertility variations in Scotland: actual, expected and ideal fertility,
CRFR Research Briefing 32, CRFR: University of Edinburgh, Edinburgh.
Watkins, S.C. (1993) If all we knew about women was what we read in demography, what would
we know? Demography, 30(4): 551–577.
Chapter 3
Women’s Education and Childbearing:
A Growing Divide
Sarah Smith and Anita Ratcliffe
Introduction
Women’s family and working lives have changed enormously over the last 25 years
in the United Kingdom (UK). Most of the changes are well-documented and several
have been discussed in other chapters – women are increasingly delaying childbear-
ing and more are remaining childless (see Simpson, 2009); they are also delaying
partnership, increasingly choosing cohabitation instead of marriage, and a grow-
ing number are raising children as lone mothers; and women are working more,
both before and after childbearing (see Hansen et al., 2009). What is shown in
this chapter is that these changes have disproportionately affected better-educated
women1 and that over time there has been an increased divergence in family and
working lives between women who have post-compulsory education and those
who do not.2
Education has long been seen as a key factor associated with patterns of employ-
ment and childbearing for women. Having a high level of education is typically
associated with later and less childbearing and the last 25 years have seen a sub-
stantial rise in levels of education among women. This is illustrated in Fig. 3.1
which shows current highest academic qualifications by cohort. Comparing the
cohort of women born in 1944–48 with that born in 1974–78, the proportion with
no academic qualifications fell from 30% to 5% while the proportion with a degree
increased from 13% to 26%. The increase in education across cohorts has been
greater for women than for men and the education gap has been substantially
narrowed.
S. Smith (B)
Department of Economics and Centre for Market and Public Organisation,
University of Bristol, Bristol, UK
e-mail: sarah.smith@bristol.ac.uk
1 For a discussion of changes in employment and childbearing among college graduates in the US,
see Goldin (2006).
2 See also Joshi (2002).

42 S. Smith and A. Ratcliffe
Proportion with qualifications

1
School Nursing/teaching
Other higher Degree
0.8
0.6
0.4
0.2
0
Male Female Male Female Male Female Male Female
1944–48 1954–58 1964–68 1974–78
Fig. 3.1 Highest educational qualification, by date of birth cohort

(Source: Authors’ calculations using data from the British Household Panel Survey, 1991–2005)
School qualifications include A levels, O levels, GCSEs and equivalents; Nursing includes
SEN, SRN and SCM, Teaching excludes degree; Other higher includes City & Guilds Cer-
tificate, HNC, HND, BEC/TEC/BTEC, Higher Certificate/Diploma, University Diplomas and
any other technical, professional or higher qualifications; the omitted category is no academic
qualifications
On its own this increase in education would have been likely to cause changes
in employment and fertility across cohorts.3 However, there have also been impor-
tant changes in family and working lives within education groups. The result has
been an increasing divergence between the experiences of women who go on to fur-
ther education and those who leave school at the compulsory school-leaving age.
This is the focus of this chapter. The next section discusses the relationship between
education and childbearing and possible reasons why that relationship might have
changed over time. We then discuss the data and present trends by education in
childbearing, employment and partnership. We conclude with a discussion of the
implications of the increasingly strong association between education and child-
bearing – for the women themselves and for children born to mothers with different
levels of education.
Education and Childbearing
Higher levels of education have typically been found to be associated with later and
lower childbearing. For the UK, Rendall and Smallwood (2003) used data from the
Longitudinal Study to compare childbearing among women with post-school quali-
fications to childbearing among women with no qualifications in a cohort of women
born between 1954 and 1958. They found for this cohort that a higher level of
education was associated with a greater incidence of childlessness (22.5% of those
3 Thiscategorization of education is fairly broad. Other studies have recognized that the field of
study may matter as well as the level of education (see Hoem et al., 2005).
3 Women’s Education and Childbearing: A Growing Divide 43
with higher qualifications remained childless compared to 15.2% of those without)

and with later entry into motherhood (half of those with no higher educational qual-
ifications had their first child by age 24, compared to one tenth of educated women).
Conditional on age at first birth they found women with higher qualifications were
more likely to go on to have a second birth and that the gap between first birth
and second birth was typically shorter. Their analysis did not look at the combined
effect of delayed first birth and higher and faster conditional progression to second
and subsequent births on completed family sizes.
Other studies however have found a negative association between women’s level
of education and completed family size. Shkolnikov et al. (2007) calculate com-
pleted family sizes by education for a number of countries (excluding the UK).
Their findings for the US and selected Western European countries, summarized in
Table 3.1, reveal the inverse relationship. Women with higher levels of education
have fewer children on average, with the biggest gap in the US and the smallest in
Sweden.
This observed negative relationship between education and childbearing may be
linked to a number of different factors. Several studies have considered the adverse
impact of early childbearing on mothers’ completion of education; although studies
which attempt to identify the causal effect of teenage pregnancy typically find a
fairly small effect (see, for example, Kaplan et al., 2004). In addition, there are
a number of ways in which higher education might be expected to affect fertility
– whether the desired number of children, or the timing. In the first instance, the
particular difficulties of combining full-time education and childbearing because
of the absence of any formal maternity provision for students is likely to result
in a delay in childbearing. Higher levels of education are also typically associated
with higher wages and so may raise the opportunity cost of taking time out of the
labour market for own childcare which would tend to reduce the number of children
(Becker, 1981; Cigno, 1991).
A further effect of education which would work to delay the onset of childbear-
ing may operate through the shape of the expected income profile over the lifetime.
Higher levels of education are typically associated with more steeply rising earnings
profiles during the early years of employment and Happel et al., (1984) argue that,
in the presence of imperfect capital markets, educated women may delay having
children in anticipation of higher earnings in the future to cover the additional costs.
Table 3.1 Average completed fertility, women aged 35–49
Year of survey Lower education Higher education
United States 1994 2.61 1.72

France 1994 2.41 1.74
Italy 1995–96 2.04 1.40
Sweden 1992–93 2.17 1.98
Source: Shkolnikov et al. (2007).

Table 3.2 Use of childcare by working women with pre-school children
Woman’s current Formal paid Informal

highest qualification childcare (%) childcare (%)
No qualifications 17.8 75.9

School qualifications 33.6 71.2
Nursing/teaching/other 45.3 67.4
higher qualifications
Degree 63.9 49.3
Source: Authors’ calculations using data from the British Household

Panel Survey, 1991–2005.
The question asks about all uses of childcare and so the totals do not
have to sum to 100%. Formal paid child care includes nanny, mother’s
help, workplace nursery, day nursery and childminder. Informal child-
care includes spouse, partner, other relative, friend or neighbour.
However, not all of the potential effect of education is to delay or reduce child-
bearing. Higher wages make formal childcare more affordable (Ermisch, 1989): as
shown in Table 3.2, there is a strong positive relationship between a woman’s level
of education and use of formal paid childcare for pre-school children. Women with
lower levels of education rely more heavily on informal childcare from family and
friends. The ability to pay for formal childcare may reduce the importance of the
opportunity cost effect for women in better-paid jobs and allow them to combine
work and family more easily, particularly if state-provided childcare is limited.4 In
addition, employers may have better incentives to retain qualified women and there-
fore offer them more flexible work schedules.
Another effect of education is to raise permanent income which will tend to
increase the desired quantity of children (Becker, 1981), an effect that is likely to
be amplified via assortative mating and partners’ higher incomes. However, Becker
also argued that higher levels of income may increase the desired quality rather than
quantity of children, with families choosing to have smaller families and devoting
more resources to each child.
Overall, the effect of higher levels of education on fertility is a priori ambiguous.
The observed relationship may suggest that negative factors dominate but, alter-
natively, it may be driven by selection effects (women with weaker preferences
for childbearing selecting into further education) rather than by the causal effect
of higher education. If selection is important, then the increasing participation of
women in post-compulsory education over time might be expected to change fer-
tility behaviour within education groups as the composition of educated women
changes (assuming no change in the underlying distribution of preferences). In addi-
tion to this selection effect, however, there are other reasons for thinking that the
relationship may have changed over time.
4 Of course childcare arrangements may reflect, rather than determine, patterns of employment.
As discussed by Dex et al. (2005) and Wolf (2006), educated women are increas-
ingly entering careers traditionally dominated by men, and they now make up a
majority of medical students and trainee solicitors and barristers. What these careers
have in common is an extended period of training compared to traditionally ‘female’
careers such as teaching and nursing and this is likely to result in further delay to
entry into motherhood because of the difficulties of combining studying and moth-
erhood. Aside from the direct enrolment effect, however, most of the impacts of
higher education on childbearing (on opportunity cost and permanent income) work
through employment choices post-education rather than being a direct effect of
education per se.
Recent estimates suggest that the returns to a degree for women have been fairly
stable across successive cohorts (see Walker and Zhu, 2003), which may suggest
little effect on either opportunity cost or permanent income. However, the types
of careers that educated women are increasingly choosing typically have a steeper
profile of earnings growth which may work to delay childbearing further either
through an opportunity cost effect or through the smoothing mechanism proposed
by Happel et al. (1984). To illustrate this, Fig. 3.2 shows a measure of occupational
wage dispersion by education and cohort. This is obtained by calculating the stan-
dard deviation of wages of all full-time employees at the occupation level (SOC90)
using data from the Labour Force Survey and matching it to the actual occupations
of women in the sample.5 The figure shows both increasing wage dispersion with
education and increasing wage dispersion over time among degree-level educated
women. As better-educated women increasingly move into occupations with steeper
wage profiles they may have more to lose if time out of employment damages
their career prospects. Since the estimates of wage dispersion are based on pooled
cross-sectional data, the differences by education and cohort are driven solely by
occupational choices rather than any change in wage dispersion over time.
5
Standard deviation of hourly pay
4
Fig. 3.2 Wage dispersion, by
cohort and highest 3
qualification
(Source: Authors’ 2
calculations using data from
the British Household Panel 1
Survey and Labour Force
Survey, 1993–2000) 0
1954–58
1964–68
1954–58
1964–68
1944–48
1944–48
Higher qualifications include

degree and other higher
qualifications (see notes to
Fig. 3.1) No qualifications Higher qualifications
5 For further information on how this is done, see Gregg et al. (2008).
Other factors, however, may tend to work in the opposite direction, encouraging
childbearing among educated women. These include increases in the generosity of
maternity provision and the improved supply of nursery places following regulation
(see Gregg et al., 2007, for an overview). While these changes in principle affected
all women, educated women may be better placed to take advantage of improved
opportunities to combine work and family because they can more easily afford to
purchase formal childcare.
Overall, there is no reason to expect that the relationship between education and
fertility should remain constant over time. However, compared to the many stud-
ies that have looked at the relationship between education and childbearing at a
particular point in time, relatively few studies have explored whether and how this
relationship is changing. For the US, Rindfuss et al. (1996) find that the association
between education and childbearing has been growing stronger over time. Over the
period 1963–89, college-educated women saw greater declines in fertility than high
school graduates; they also experienced a greater shift towards later childbearing.
By contrast, recent work by Kravdal and Rindfuss (2007) found that, for Norway,
the education gap in fertility has narrowed across successive birth cohorts. Unlike
the US, the greatest reductions in childbearing have occurred among the less well-
educated, while childbearing among the better-educated has been relatively stable. 6
This chapter adds to this literature by looking at what has happened to the relation-
ship between education and fertility in the UK over time.
Data and Methodology
The data we use to analyse childbearing by education are drawn from successive
waves of the Family Expenditure Survey (FES) 1978–2003, supplemented with
waves of the Family Resources Survey (FRS) from 1995–2004. Both are cross-
sectional household surveys, containing a standard set of demographic and socio-
economic variables on household members, as well as detailed information on
household spending (the FES) and income (the FRS). The surveys do not specif-
ically collect information on women’s fertility histories. Instead, we use the ‘own
child method’ (see Murphy and Berrington, 1993) to infer fertility histories from the
age of the mother and the age of her natural-born children living in the household.
First, we allocate children in the household to their natural mothers using infor-
mation provided on relationships between household members and the benefit unit
to which individuals belong. Then, for each woman we create pseudo-fertility his-
tories – the age of the mother at birth and the birth order of the child – based on the
current ages of the mother and children. So, for example, a woman aged 25 who has
one child aged 0 is assumed to have had her first child at age 25; a woman aged 30
who has one child aged 2 is assumed to have had her first child at age 28; a woman
6 The exception was women with higher degrees who did see completed fertility increase over time
but this group constituted a small proportion of the sample.
aged 35 who has one child aged 10 and another aged 8 is assumed to have had her
first birth at age 25 and her second birth at age 27, and so on. As these examples
illustrate, we combine ‘current’ birth probabilities (i.e. whether or not a woman has
a birth in the year in which she is observed in the FES/FRS sample) with retrospec-
tive birth probabilities (i.e. whether she had a birth in previous years based on the
ages of her children). Finally, we use survival analysis to construct cohort parity
progression ratios from the individual fertility histories.
Clearly, the own child method is not without its potential problems. One is that
we observe the current ages of the mother and any children, but not their actual dates
of birth. In practice, a woman aged 25 who has one child aged 0 may actually have
given birth when she was 24. Since the woman could not be a year older than we
currently observe her to be when she gave birth, our estimates of the proportion of
women who have births of order b by a particular age, and the average age at birth
order b, will tend to have a systematic downward bias. However, the bias should
apply equally to all cohorts and education groups, and should not therefore affect
the validity of the inferences we draw about differences in fertility behaviour over
time and by education.
A second problem is that the own child method relies on information on surviving
children in the household to infer fertility histories – infant mortality and household
reconstitution will result in measurement error. However, low rates of child mortal-
ity7 and the fact that the overwhelming majority of children stay with their natural
mother in the event of family break-up act to reduce the effect of these factors in
practice.
Another potential problem – one that is more serious in practice – is that older
women may have had children who have now left home. If we observe a woman
aged 39 with no children, we cannot say for certain whether she has not yet had
children, or she had one child when she was 17 who left home when they were 20.8
In the latter case, the own child method would fail to capture births to women who
had children relatively young – we would tend to over-estimate the mean and median
ages associated with different birth orders (i.e. the average ages of the mother at
first, second and subsequent births) and to under-estimate family size for people
who entered childbearing at a relatively young age.
A solution to this problem is to adopt a maximum age threshold, i.e. to exclude
from our analysis women above a threshold age at which the problem of children
leaving home starts to significantly affect the estimates of the proportion of women
having births at younger ages. Assuming that women start having children from age
16, the selection problems may arise from as young as age 32. In fact, sensitivity
7 There was a decline in infant mortality over the period. The rate of death in England and Wales of
children less than one year was 14.3 per 1,000 live births in 1976, falling to 5.0 per 1,000 live births
in 2006 ”Infant and perinatal mortality 2006: health areas, England and Wales”, Health Statistics
Quarterly 35. Our estimates will therefore tend to underestimate births more at the beginning of
the period.
8 The problem is made potentially worse in practice by the fact that students who live away from
home are not counted as part of the household in the FES/FRS.
analysis of estimates of the proportion of women having a first birth by age 20 made
using successively older samples, shows that the threshold can be raised to 37 before
there are significant effects.9
Our analysis of fertility therefore excludes births after age 38. This means that
we under-estimate the proportion of each cohort having first and subsequent births.
Murphy and Grundy (2003) suggest that fewer than 1% of births are to women aged
40+ among cohorts born after 1930. However, the recent trend towards later child-
bearing suggests that the bias is likely to increase among later cohorts. Moreover,
the effect of the bias will not be the same across education groups. This must be
borne in mind when interpreting our results and we place more emphasis on our
analysis of the age at first birth rather than completed fertility.
The advantages of generating cohort fertility profiles by applying the own child
method to repeated cross-section data from the FES and FRS are that we have
full information on birth order, which is only recently available in complete form
in data derived from official birth registration data. We also have information on
the mother’s education, as well as other economic and demographic characteristics
which we use to document changes in employment and partnership among women
with different levels of education.
As a check of the validity of our approach, we compare an estimate of the period
total fertility derived from the FES/FRS data from 1968 onwards, with the official
measure of total fertility derived from registration data, shown over a longer period
to highlight fertility trends. Total fertility measures the number of children a woman
would have if she experienced the age-specific fertility rates in that year. As shown
in Fig. 3.3, total fertility estimated using the FES/FRS is very close to the official
measure. As is to be expected, our estimate is lower than the official measure since
we exclude births over 37. However, we pick up the major trends in fertility (in par-
ticular, the decline in total fertility from around 2.5 in 1968 to 1.7 by the end of the
period), and the average difference is relatively stable over time.
3.5
ONS FES/FRS
3
Total fertility rate
2.5
1.5
Fig. 3.3 Total fertility – 1
FES/FRS estimate and ONS
estimate 0.5
(Social Trends 30 and Series 1930 1940 1950 1960 1970 1980 1990 2000
FM1 No.35) Year
9 This suggests that there may be a negative relation between the age of the mother at first birth and
the age at which the child leaves home.
Changes in Childbearing by Education

Rendall and Smallwood (2003) looked at differences in childbearing in women born
in 1954–58, between those with and without post-secondary-school qualifications.
We extend the analysis to consider women born in earlier and later cohorts in order
to assess the extent to which the education gap has changed over time. We focus on
women born ten years’ earlier (i.e. 1944–48) and ten years’ later (i.e. 1964–68).
Information on academic qualifications is not available in all waves of the FES
and FRS. Instead, there is consistent information available since 1978 on the age at
which people leave full-time education. Replicating the exact split made by Rendall
and Smallwood is therefore not possible. Instead, we define two groups of edu-
cated women – one who left full-time education at 19+ (we call these women with
higher qualifications) and the other who left full-time education at 21+ (we call
these graduate women). In both cases, we compare their outcomes with those of
women who left school at 16 or below (we call these school leavers). Rendall and
Smallwood report that 17% of their sample had obtained a higher qualification; the
figures reported in Table 3.3 below suggest that this corresponds closely to the pro-
portion leaving full-time education at age 19+. However, compared to Rendall and
Smallwood, our sample may differ in that we may miss some women who obtained
academic qualifications after they left full-time education, and not take account of
women who stayed on in full-time education but did not obtain any qualifications.
Table 3.3 presents two sets of summary statistics on childbearing by cohort and
education – completed fertility and the proportion remaining childless. In both
cases, as already discussed, we only include women up to age 38. Rendall and
Smallwood do not calculate completed fertility by education, but we can com-
pare our estimates of childlessness with theirs for the 1954–58 cohort. For women
leaving school at 16 (corresponding to their sample of women with no educational
Table 3.3 Summary measures of childbearing, by cohort and education
1944–48 1954–58 1964–68
Completed fertility up to age 38

Left school at 16 1.97 1.93 1.94
Left FT education at 19+ 1.58 1.45 1.40
Proportion childless up to age 38
Left school at 16 0.125 0.144 0.149
Proportion of cohort
Source: Authors’ calculations using data from the Family Expenditure

Survey and Family Resources Survey, 1978–2005.
qualifications), we find that 14% are childless, compared to 15% in the Longitudinal
Study. For women with higher qualifications, however, our estimate of childlessness
is quite a bit higher – nearly 30% compared to 22.5% in the Longitudinal Study. Of
course, births after the age of 38 may account for the difference. We obtain a much
closer match when we look at the proportions having a first birth at different ages.
In the Longitudinal Study, 10% of women with higher qualifications had their first
birth by age 24 compared to 50% of women without. We obtain figures of 10% and
53% respectively.
The figures in Table 3.3 illustrate that the divide in childbearing by education
appears to have been getting wider. Better-educated women in the 1944–48 cohort
had lower completed fertility and a higher level of childlessness than those who
left school at 16. By the 1964–68 cohort this gap had widened. Completed fer-
tility fell by more among the group with higher qualifications than among those
who left school at 16 and the gap in completed fertility between the two groups
increased from 0.42 to 0.58. Of course, we cannot rule out that higher fertility at
later ages among educated women may be enough to maintain levels of childbear-
ing but other studies based on the UK cohort studies also point to increasing levels
of childlessness among educated women (see Kneale and Joshi, 2007). The Table
suggests that most of the increase in the gap occurred between the 1944–48 cohort
Left school at 16
0.5
Proportion of women
0.4
0.3
0.2
0.1
0
<20 20–24 25–29 30–34 35+
Age
1944–48 1954–58 1964–68 1974–78
Left full-time education at 21+

0.5
Proportion of women
0.4
0.3
Fig. 3.4 Age at first birth, by 0.2

cohort and education
(Source: Authors’ 0.1
the Family Expenditure 0
Survey and Family Resources <20 20–24 25–29 30–34 35+
Survey 1978–2005) Age
and the 1954–58 cohort; changes between the later cohorts appear to have been
smaller.
The increased divide between education groups is clear if we look at changes in
the timing of first birth – shown for different date-of-birth cohorts in Fig. 3.4. For
women leaving school at age 16, there has been a decline in the size of the peak at
ages 20–24, indicating a slight increase in the age at first birth. However, the peak
has remained at 20–24 and there has been very little increase in the proportion of
women having their first birth in their late 20s or 30s. There has been an increase in
the proportion of women leaving school at 16 who are teenage mothers, particularly
among the most recent cohort of women (born 1974–78).
By contrast, there has been a clear rightward shift in the distribution of age at
first birth among graduates. Even among the oldest cohort, childbearing began later
than among women leaving school at 16, but the most common age of entry into
motherhood for graduates was 25–29 and most had their first child before age 30.
By the 1964–68 cohort, this was no longer the case with the peak shifting to 30–34.
Early evidence from the 1974–78 cohort however, shows less evidence of a further
big shift in the age distribution.
The result is a greater polarization in age of childbearing by education. This is
clearly illustrated in Fig. 3.5 which shows, for the 1944–48 and 1964–68 cohorts,
the proportion of women having a first birth before 30 according to the age at which
Cohort born 1944–48

0.8
Proportion of women
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
16 17 18 19 20 21
Age completed full-time education
Proportion of cohort First birth before 30
Cohort born 1964–68

0.8
Proportion of women
0.7
0.6
Fig. 3.5 Proportion with first 0.5
birth by age 30, by cohort and 0.4
education 0.3
(Source: Authors’ 0.2
calculations using data from 0.1
the Family Expenditure 0
Survey and Family Resources 16 17 18 19 20 21
Survey 1978–2005) Age completed full-time education
they left full-time education. It highlights that the educational gradient steepened as
educational attainment increased. In the earliest cohort the majority of all women,
regardless of their education, had their first child by the time they were 30; this is no
longer the case. There has been very little change in the proportion of women leav-
ing school at 16 who have their first child by age 30 which remains over 70%. But
there have been big changes for educated women – among those leaving full-time
education at 21+ the proportion entering motherhood by age 30 has halved from
around 60% to 30%. In the next section, we show that these changes in childbearing
have been accompanied by differential changes in employment and partnership for
educated women.
Changes in Employment and Partnership, by Education
Information on employment and partnership by education are available only in the

years of the survey (1978–2004) and unlike the case of fertility we cannot construct
retrospective histories; we therefore observe only parts of the age profiles for each
of the cohorts using standard pseudo-cohort analysis. These are shown in Figs. 3.6
and 3.7. Each of the lines represents the age profile of a single cohort; gaps between
the lines therefore indicate changes in behaviour across cohorts compared at the
same˜age.
Figure 3.6 shows age profiles of average weekly hours. Weekly hours are zero
for those not working and the changing profiles therefore reflect changes in partic-
ipation as well as shifts between full-time and part-time employment. For school-
leavers there has been an increase in average hours across cohorts but this has been
fairly small and has caused little change in the overall shape of the age profile – the
traditional ‘u’ dipping and rising at around the same ages, consistent with the peak
in age at first birth remaining at 20–24.
The sample size of graduates in each cohort at each age is relatively smaller than
that of school leavers, reflected in less smooth profiles. Nevertheless, the main trends
are clear. There has been an increase in hours worked by graduate women in their
20s and 30s across successive cohorts and the shape of the age profile is moving
Left 16 Left 21+

40
1944–48 1954–58 1964–68 1974–78
Mean hours per week
30
Fig. 3.6 Age profile of
weekly hours worked, by
cohort and education 20
(Source: Authors’
10
the Family Expenditure
Survey and Family Resources 20 30 40 50 20 30 40 50
Survey 1978–2005) Age
Fig. 3.7 Employment among Youngest child <5

0.8
women with pre-school <30 hours 30+ hours
children, by cohort and
Proportion working
education 0.6
(Source: Authors’
0.4
Survey and Family Resources
Survey 1978–2005) 0.2
0
Left 16 Left 21+ Left 16 Left 21+ Left 16 Left 21+
1944–48 1954–58 1964–68
away from the traditional ‘u’, in part reflecting an increasing dispersion of age of
entry into motherhood. There have also been changes in employment before and
after childbearing. First, there has been a prolonged period of full-time employment
prior to childbearing with very little decline in average weekly hours before age 30
for both the 1964–68 cohort and the 1974–78 cohort. Second, there has also been an
accelerated return to work after childbearing. Using data from the UK birth cohorts,
Joshi (2002) reports that the median return to work time for graduate mothers fell
from five years in the 1970s to less than one year in the 1990s. For women with no
qualifications the median return to work time remained at five and a half years. The
result is an increase in employment among women with young children, particularly
among graduates. This is shown in Fig. 3.7.
Between the 1944–48 cohort and the 1964–68 cohort, the rate of employment
of graduate women with pre-school children increased from 40% to 60% and
almost all of the increase was from a rise in full-time employment (>30 hours).
For women who left school at 16, the increase in the rate of employment was
Left 16 Left 21+

1
Proportion in a couple
0.5
1944–48 1954– 58
1964–68 1974– 78
0
20 30 40 50 20 30 40 50
Age
Fig. 3.8 Age profile of partnership, by cohort and education

(Source: Authors’ calculations using data from the Family Expenditure Survey and Family
Resources Survey 1978–2005)
smaller (from 26% to 43%) and was driven by rises in both part-time and full-
time employment. By the 1964–68 cohort, part-time employment rates among
school leavers had caught up with those of graduate mothers but the full-time
employment rate was still much lower. This is consistent with the pattern of
childcare use among the two groups with educated women making more use
of formal paid childcare which is more likely to be available on a full-time
basis.
Among graduate women the biggest increase in full-time employment occurred
between the 1944–48 and 1954–58 cohort, most likely reflecting the introduction
of maternity leave in 1975 which benefited the later cohort. This has allowed more
women to combine employment and children as shown by the increase in full-time
employment among women with pre-school children (shown in Fig. 3.7). It may
also explain the move towards “career” jobs since maternity leave allows women
to combine work and family, enabling them to reap some of the benefits of early
investment in a career.
Trends in partnership have also differed by education. For graduate women, there
has been a trend towards delayed partnership formation (here partnership includes
both formal marriage and cohabitation) but the eventual rate of partnership among
graduate women in their mid-30s has remained fairly stable at around 80%. For
women leaving school at age 16, the story is quite different. Not only are part-
nership rates lower at younger ages, but they have shifted downwards at all ages
with each successive cohort. The rate of partnership at age 35 has fallen from 85%
among the 1954–58 cohort to below 70% among the 1964–68 cohort. There has
been a correspondingly greater increase in lone-motherhood among this group.
Among graduate women with pre-school children, the proportion who are lone
mothers increased from 2% among the 1944–48 cohort to 5% among the 1964–
68 cohort. Among those leaving school at 16, the increase was from 6% to 22%
(Fig. 3.9).
Proportion who are a lone mother
0.4
Youngest child <5 Youngest child 5–10
0.3
0.2
0.1
Fig. 3.9 Lone-motherhood,
by cohort and education
(Source: Authors’ 0
1944–48
1954–58
1964–68
1944–48
1954–58
1964–68

Survey and Family Resources
Survey 1978–2005) Left 16 Left 21+
Conclusions and Discussion

By comparing childbearing, employment and partnership across successive date of
birth cohorts, this chapter has highlighted the increased polarisation between the
family and working lives of women with different levels of education.
Across cohorts of women born in 1944–48, 1954–58 and 1964–68, the pattern
of childbearing and employment for women leaving school at 16 has remained
relatively stable. There has been little change in the age at which they begin child-
bearing; around half have their first child by age 25 and more than three-quarters
by age 30. There has been an increase in employment among those with children,
but part-time work remains the norm. The biggest change has been in partner-
ship, with an increasing proportion of lone mothers. This reflects both increas-
ing rates of divorce and separation and a rise in the number of never-married
mothers.
Graduate women have experienced much greater changes in childbearing and
employment. Childbearing is increasingly delayed and fewer than 30% now have
their first birth before age 30. There has been a greater increase in full-time employ-
ment – both before and after employment. Partnership has been delayed but whilst
the eventual rate of partnership remains much the same, the rise in lone-motherhood
has been smaller. Childbearing and partnership are put on hold as more graduate
women pursue careers but more (if not the majority) now combine careers with
motherhood.
Thus increasing levels of educational attainment across successive generations
of women can account for some – but not all – of the observed changes over time in
women’s family and working lives. The lives of graduate women themselves have
also changed compared to the lives of earlier cohorts of graduate women. Predicting
what will happen in the future to women’s childbearing and employment there-
fore involves more than just observing changes in educational attainment, but also
understanding the evolving family and working lives of women at different educa-
tion levels. For later cohorts (born 1974–78) the evidence presented here points to
less dramatic shifts in behaviour among graduate women. Figure 3.4 shows similar
rates of childbearing by age 30 to the earlier cohort, while Table 3.4 shows that rates
of employment among graduate women with pre-school children have not continued
to increase at such a fast rate. This may indicate that, while there is now a greater
divide, it is less likely to widen further.
The increased polarisation is likely to have hugely important effects for the lives
of the women involved and for their children. The fact that women of different
education backgrounds no longer share recognisably the same family and working
lives is seen by some as marking the end of a common sisterhood (Wolf, 2006).
For children, the changes in work and family lives mean a greater material advan-
tage for those born to educated mothers. Table 3.4 summarises information on
maternal employment, lone-motherhood and household income for pre-school chil-
dren (<5) by mother’s education and year. Unlike previous analysis, which com-
pared women of different date-of-birth cohorts whose children would have been
Table 3.4 Women with pre-school children, by education and year
Proportion working Proportion working Proportion who Mean real weekly

30+ hours < 30 hours are lone mothers income
16 21+ 16 21+ 16 21+ 16 21+
1978–82 .048 .063 .198 .272 .084 .026 £55 £79

1983–87 .075 .143 .221 .333 .124 .023 £73 £121
1988–92 .095 .242 .253 .295 .183 .037 £87 £181
1993–97 .114 .304 .282 .312 .254 .065 £115 £210
1998–02 .137 .315 .318 .317 .282 .069 £139 £231
Source: Authors’ calculations using data from the Family Expenditure Survey and Family
Resources Survey 1978–2005.
Income is household disposable income in constant 2003 prices.
born at different points in time because of differences in childbearing ages, this

directly compares children of the same age at the same point in time.
Of course, the picture from looking at trends in employment and lone-
motherhood over time is much the same as looking across cohorts. However, the
final column shows a measure of household income. The aim is to look at what has
happened over time to family material well-being for children born to women with
different levels of education. Reflecting the changes to age at first birth, maternal
employment and partnership, there has been an increasing divide between house-
hold incomes, by education. At the beginning of the period (1978–82) the ratio
of household incomes of female graduates to school leavers was 1.43. Ten years
later, this had grown to over 2. More recently, the gap has narrowed back to 1.66,
likely as a result of large increases in child-contingent benefits targeted mainly at
low-income households. Between 1999 and 2003, government spending per child
increased by 50% in real terms, a change that was unprecedented over the previ-
ous thirty year period and most of the additional spending was targeted at low-
income households. However, even this massive increase in government spending
has not been enough to counter the effect of changes in employment, childbearing
and partnership on relative incomes. Compared to 25 years ago, the income gap has
widened.
Children born to graduate women now live in households with relatively higher
incomes in their early childhood. Because of rising levels of full-time employ-
ment, however, they are also more likely to spend time in formal childcare at
young ages. Reductions in inputs of maternal time may offset the positive impact
of increases in household income. Predicting how these changes will play out in
terms of later outcomes depends on the overall effects of both these changes. There
are extensive literatures which attempt to measure the causal effect of income on
child outcomes (health and educational attainment) and of maternal employment on
child outcomes. Very broadly these literatures suggest that a higher level of per-
manent income has a positive effect, particularly at younger ages, although this is
relatively small compared to the effect of family background variables, including
maternal education (Blow et al., 2005). Maternal employment has generally not
been found to have a negative effect (Gregg et al., 2005), with some evidence for
long-term positive effects of pre-school formal childcare (Goodman and Sianesi,
2005). If anything, therefore, the changes in working and family lives that have
occurred seem likely to amplify further the advantages for children born to graduate
women.
References
Becker, G. (1981) A Treatise on the Family, Harvard University Press, Cambridge Mass.
Blow, L, Goodman, A, Walker, I. and Windmeijer, F. (2005) Parental Background and Child Out-
comes: How Much Does Money Matter and What Else Matters? Department for Education and
Skills Research Report RR660 http://www.dfes.gov.uk/research/data/uploadfiles/RR660.pdf
Cigno, A. (1991) Economics of the Family, Clarendon Press, Oxford.
Dex, S, Ward, K. and Joshi, H. (2005) Changes in women’s occupations and occupational mobility
over 25 years, CLS Working Paper, Centre for Longitudinal Studies, London.
Ermisch, J. (1989) Purchased childcare, optimal family size and mother’s employment; theory and
econometric analysis, Journal of Population Economics, 2(2): 79–102.
Goldin, C. (2006) The quiet revolution that transformed women’s employment, education and fam-
ily, American Economic Review, 96(2): 1–21.
Goodman, A. and Sianesi, B. (2005) Early education – how long do the impacts last? Fiscal Studies,
26(4): 513–548.
Gregg, P., Burgess, S., Propper, C. and Washbrook, E. (2005) The effects of a mother’s return to
work decision on child development in the UK, The Economic Journal, 115: F4–F80.
Gregg, P. Gutierrez-Domenech, M. and Waldfogel, J. (2007) The employment of married mothers
in Great Britain, 1974–2000, Economica, 74(296): 842–864.
Gregg, P., Grout, P., Ratcliffe, A., Smith, S. and Windmeijer, F. (2008) How important is pro-social
behaviour for the delivery of public services? CMPO Working Paper, Bristol.
Happel, S., Hill, J. and Low, S. (1984) An economic analysis of the timing of childbirth, Population
Studies, 38: 299–311.
Hansen, K., Hawkes, D. and Joshi, H. (2009) The timing of motherhood, mothers’ employment
and child outcomes, in Stillwell, J., Coast, E. and Kneale, D. (eds.) Fertility, Living Arrange-
ments, Care and Mobility: Understanding Population Trends and Processes Volume 1, Springer
Dordrecht, pp. 59–80.
Hoem, J, Neyer, G. and Andersson, G. (2005) Childlessness and educational attainment among
Swedish women born 1955–59, WP-2005–014, Max Planck Institute for Demographic
Research, Rostock.
Joshi, H. (2002) Production, Reproduction and Education: Women, Children and Work in a British
Perspective, Institute of Education mimeo.
Kaplan, G., Goodman, A. and Walker, I. (2004) Understanding the effects of early motherhood in
Britain: the effects on mothers, Discussion Paper No. 1131, Institute for the Study of Labor
(IZA).
Kneale, D. and Joshi, H. (2007) Social polarisation in the timing of first birth, Presentation to
British Society of Population Studies, University of St Andrews, September 12th 2007.
Kravdal, O. and Rindfuss, R. (2007) Education and fertility in Norway – a study of how the rela-
tionship has changed over time among women and men, Presentation to GeNet Seminar ‘Low
Fertility in Industrialised Countries’, London School of Economics, May 31st 2007
Murphy. M. and Berrington, A. (1993) Constructing period parity progression ratios from house-
hold survey data, in Ni Bhrolchain, M. (ed.) New Perspectives on Fertility in Britain, Stud-
ies on Medical and Population Subjects No. 55, Office of Population Censuses and Surveys,
London.
Murphy, M. and Grundy, E. (2003) Mothers with living children and children with living mothers:
the role of fertility and mortality in the period 1911–2050, Population Trends, 112: 36–48.
Rendall, M.S. and Smallwood, S. (2003) Higher qualifications, first-birth timings and further child-
Rindfuss, R., Morgan, P. and Offutt, K. (1996) Education and the changing age pattern of American
fertility, Demography, 33(3): 277–290.
Simpson, R. (2009) Delayed childbearing and childlessness, in Stillwell, J., Coast, E. and Kneale,
D. (eds.) Fertility, Living Arrangements, Care and Mobility: Understanding Population Trends
and Processes Volume 1, Springer Dordrecht, pp. 23–40.
Shkolnikov, V., Andreev, E., Houle, R. and Vaupel, J. (2007) The concentration of reproduction
in cohorts of women in Europe and the United States, Population and Development Review,
33(1): 67–99.
Walker, I. and Zhu, Y. (2003) Education, earnings and productivity: Recent UK evidence, Labour
Market Trends, 111(3): 145–152.
Wolf, A. (2006) Working girls, Prospect Magazine, (April).
Chapter 4
The Timing of Motherhood, Mothers’
Employment and Child Outcomes
Kirstine Hansen, Denise Hawkes and Heather Joshi
Introduction
The last decades of the twentieth century have witnessed, as part of the second
demographic transition affecting most industrial countries (Van de Kaa, 1987), two
features of particular interest here – a delay in entry to motherhood and an increased
chance of women being employed even after they have children. In Britain, these
trends have not been experienced uniformly across the social spectrum. The research
reported in this chapter attempts to establish links between these two phenom-
ena, to quantify the extent of social differentials (in the timing of first motherhood
and maternal employment) and to investigate their impact on the development of
children.
In the not so distant past, very few mothers had paid work, especially during
the early years of their child’s life. Becoming a mother in your early twenties and
staying at home was the normal course of events and the conventional wisdom at
the time disapproved of anything, such as employment, which might disrupt the
bond (or attachment) between mother and her child. However, over the past fifty
years much has changed. There was a baby boom in the 1950s and 1960s followed
by falling fertility rates and rising age at first, and subsequent, births from 1970
onwards. By the twenty-first century, mothers working had become the norm. Two
thirds (65.4%) of mothers with dependent children were in employment in 2002.
Over half (52.9%) of those with a child under five had jobs (Duffield, 2002). Fifty-
four per cent of the members of the Millennium Cohort Study had an employed
mother when they were three years old (at sweep two, 2003–5) whilst 58% had an
employed mother when they were five years old (at sweep three, 2006). Of those
mothers who were employed at sweep two, 76% were working part-time (Ward
and Dex, 2007). However, the increase in maternal employment has not occurred
K. Hansen (B)
e-mail: K.Hansen@ioe.ac.uk

60 K. Hansen et al.
uniformly across the population. Those who are more likely to be in paid work are
also those who are most likely to have had their children relatively late. Indeed, the
age at which a woman enters motherhood has become more spread out in recent
years, with those from the most disadvantaged situations still tending to become
mothers at a young age, and subsequently, being most likely to be out of the labour
market.
For the families of mothers in employment, a key decision is who takes respon-
sibility for their children whilst they are at work. There are social differences
here too, with those who can afford it purchasing formal childcare from trained
providers, while for others cost remains a barrier to this type of care. Other parents
put a premium on informal care, attracted by knowing the carer or the flexibility
of making informal arrangements with family and friends. The impact of these
childcare modes on children is a growing area of research amongst social scien-
tists (UK studies include Gregg et al., 2005; Hansen and Hawkes, 2009; Sylva et al.,
2004).
Since 1997, the Government has brought in a suite of measures to reduce child
poverty and inequality. The Social Exclusion Unit (SEU) has initiated attempts
to reduce teenage motherhood and to attempt to improve life for the mother as
well as the child (SEU, 1999). Several other policies have attempted to reduce
child poverty by increasing the employment of mothers. These include the pro-
vision of good quality childcare, local child and family-based initiatives through
the Sure Start programme, the introduction to the right to ask for flexible work-
ing agreements, the extension of maternity leave, the introduction of paternity leave
and, for those who have been out of the labour market for some time, the intro-
duction of additional training and job search advice through ‘New Deal for Lone
Parents’.
This drive towards increasing mothers’ employment, especially when children
are very young, has led many researchers, as well as the public, to ask about the
impacts on the children. Our research has focused on the consequences of maternal
employment and the timing of motherhood for cognitive and behavioural outcomes
in children. Our data resource is the UK birth cohort studies of 1958, 1970 and the
Millennium (Bynner and Joshi, 2007). We use these data to illuminate differences
across families by age at first motherhood, maternal employment, childcare and
subsequent child outcomes.
This chapter therefore presents the results of our project in three key areas: the
timing of motherhood, maternal employment, childcare and the consequences for
the children. We address the following questions:
• What are the main determinants of age at motherhood, and have these changed
between the cohorts?
• What are the main determinants of maternal employment in a child’s early life?
• What are the impacts of the childcare decisions, maternal employment and age
at motherhood on cognitive and behavioural outcomes for children?
4 The Timing of Motherhood, Mother’s Employment and Child Outcomes 61
In the conclusion, we turn to the possible impact of Government policies in

the last ten years on child cognitive and behavioural outcomes by addressing the
following questions:
• Will the policies of the last decade help to reverse the increasing polarisation of
age at motherhood and maternal employment that we saw develop in the second
half of the 20th century?
• Will this help to reduce the social disadvantage felt by those who grow up in
some of the most socially disadvantaged households in the UK?
In undertaking this research, we are adding to an existing literature on the socio-

economic variations around the timing of the initiation of childbearing, which it may
be useful to review briefly.
Literature on Timing of Motherhood

Despite a general and international trend towards later motherhood, early child-
bearing is still relatively common in the UK compared to other European countries
(Chandola et al., 2002). It is also distinctive for its socio-economic covariates
(Buxton et al., 2005; Rendall et al., 2008). Less advantaged women still tend to
have children earlier and it is the most advantaged who have increasingly deferred
childbearing (Rendall et al., 2005). This social polarisation in age at motherhood has
emerged in Britain as the trend towards later childbearing in the last three decades
of the century has affected various social groups differentially, and has differential
consequences.
There are many studies of the socio-economic determinants of the timing of
motherhood (e.g. Kiernan and Diamond, 1983; Kiernan, 1992, 1997; Ermisch and
Pevalin, 2003a; Rendall and Smallwood, 2003). In particular, teenage motherhood
is associated with exclusion from both employment and education (Bynner and
Parsons, 1999; SEU, 1999). Research attempting to disentangle the causes of early
motherhood and associated problems suggests that early motherhood is a marker
rather than a driver of subsequent labour market disadvantages. Work by Ermisch
and Pevalin (2003b, 2005) on the British birth cohort of 1970 suggests that a teen
birth has little independent effect on a woman’s qualifications, employment or earn-
ings at age 30. Work on sisters in the US (Geronimus and Korenman, 1992) and
twins in the UK (Hawkes, 2008) also suggest that early motherhood is strongly
associated with poor family background. The apparent effects of entering mother-
hood early on household income and educational attainment are much smaller once
antecedent factors are taken into account. However Ermisch and Pevalin (2003b,
2005) do attribute the lower employment and educational attainment of any partner
present when a teen mother reaches 30 to a reduction of prospects in the ‘marriage
market’. Liao (2003) finds another independent, adverse, effect of early motherhood
on a woman’s mental health.
62 K. Hansen et al.
The timing of motherhood, early or late, may be the outcome of a deliberate strat-
egy or it may be unintended, early because of unplanned pregnancy, or late because
of unplanned infertility (or unanticipated absence of a suitable partner). Probably
each type of account applies to some individuals. Early motherhood may constitute,
for some, a rational choice in the face of limited alternative prospects in education
or employment, along with the safety net of the UK benefit system (Rendall et al.,
2008). For others, the main explanation might be a lack of planning and informa-
tion (Barrett and Wellings 2002; Allen and Bourke Dowling, 1998; Kneale, 2009).
Young mothers may or may not have positively wanted to have a baby at the time
they did, and they may not be consciously aware of the state benefits available.
Whatever the explanation for their mothers’ behaviour, any differential outcomes
for families started early and late are certainly not chosen by the children. Whatever
its cause, young age at motherhood could be viewed as a signal of disadvantage on
a number of fronts, including consequences for the children (Pevalin, 2003).
The implications of these associations are not only present for the mother, but
also for the next generation. As the established British Birth cohort studies of 1958
and 1970 have shown (Gregg et al., 1999; Hobcraft and Kiernan, 2001; Feinstein,
2003; Blanden et al., 2005; Schoon, 2006), being born into an advantaged or disad-
vantaged family may affect the experience of childhood and prospects in later life
for children. Being born to a mother who had her first child in her teens or early
twenties is emerging as an increasingly distinctive feature of the intergenerational
transmission of disadvantage, which at the same time accelerates the cycle.
Timing of Motherhood
Although the post-war era started with rising fertility and falling age of childbearing,
the decades since 1970 and the generations of women born since 1940 have been
characterised by a general postponement of motherhood. Later first birth pushes
second and subsequent births to later ages, although there has also been a downward
trend in the proportions proceeding to families larger than two. The age by which
half of a cohort had embarked on having at least one child reached a low of around
23 for cohorts of women born in the 1940s and has been rising ever since to nearly 29
for cohorts born in the mid 1970s. These estimates for England and Wales are based
on births up to 2006 and the adjustments made for true birth order (Smallwood,
2002; ONS, 2007, Table 10.3).
An alternative indicator is the mean age of women having their first birth in a
given calendar year. This statistic was 27.6 years in 2006 and 25.5 in 1990, as shown
in Fig. 4.1. This period measure of age at entry to motherhood is plotted for the years
1990–2006 for all birth orders, and first births only, incorporating adjustments for
true birth order (Smallwood, 2002; ONS, 2007, Table 1.7b). This shows that across
those 16 years the age at first birth rose by nearly two years, while the age of mothers
at all births rose by 1.4 years. The upward drift of delayed entry to motherhood also
pushes up ages at higher orders.
Fig. 4.1 Mean age at all and 30

All births First births (true order)
first births, England and 29
Wales, 1990–2006
28
Mean age
(Source: ONS, 2007;
Smallwood, 2002) 27
26
25
24
23
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
Year
Figure 4.2 shows a time series for England and Wales from 1958 until 2004
based on the more generally available data on first births within marriage, and all
births outside marriage, for which birth order is not collected at registration (ONS,
2005). We can see that since 1970 for both groups there has been an upward trend in
the age at birth. We can also see that those who had births outside of marriage were
younger, on average, than those who had first births inside of marriage, despite the
rise in childbearing within cohabitations boosting the numbers of non-marital births
in the later years.
Table 4.1 presents the mean age at childbearing in three UK birth cohort studies
and the national registration taken from the ONS (2005) for first births and all
births. For both series, the age of the cohort child’s mother in the three cohort
studies reflects the general trend found in the national data, with ages starting to
rise from 1970. Mothers in the Millennium Cohort Study (MCS) were significantly
older, in terms of both all birth orders and first births only, than those in the
National Child Development Study (NCDS) of 1958 as well as the British Cohort
Study (BCS) in 1970. The difference in age at first births is nearly four years (3.8)
between 1970 and 2000–1.
The upward trend in the age at first birth has not applied uniformly to the whole
distribution, which is more easily seen by considering cohorts, women themselves
born in a particular year, rather than as above, all the women giving birth in a
particular year. Figure 4.3 presents the age by which successive quintiles of the
32
Births outside marriage First births within marriage
30
Mean age
28
26
24
Fig. 4.2 Average age at first 22
births within marriage and all
births outside marriage, 20
1958
1960
1962
1964
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1988
1990
1992
1994
1996
1998
2000
2002
2004
England and Wales,

1958–2004
(Source: ONS, 2005). Year
64 K. Hansen et al.
Table 4.1 Age of mothers at first and all births, three birth cohort studies and registration statistics
NCDS BCS MCS

Year 1958 1970 2000–1
Mean age of mothers 24.4 23.1 26.9

having first birth
[95% Confidence Interval] [24.3, 24.5] [22.9, 23.2] [26.5, 27.2]
Mean age of all mothers at 27.5 26.0 28.9
birth of cohort
[95% Confidence Interval] [27.4, 27.5] [25.9, 26.1] [28.7, 29.1]
Births of any order 27.8 26.2 29.1
registered in England and
Wales (ONS, 2005)
Source: Authors’ analysis; ONS (2005).
whole cohort had entered motherhood. This shows that the first 20% of successive
cohorts have entered motherhood under or around age 20. The rising age of
motherhood for cohorts born since 1940 has particularly affected ages over 25 with
half of the cohort entering motherhood later, if at all. The number of years between
the first 20% and the last was 8 years for the 1940 cohort and 18 for the last one we
have a complete fertility history for, those born in 1965. In the 1940 cohort entry
to motherhood was relatively compressed into the twenties. Two to three decades
later it is becoming more spread out, and as we shall see below, more socially
differentiated.
We have already seen in Fig. 4.2 that the persistence of young entry to mother-
hood is associated with births outside marriage. Another dimension, along which
women vary by age at first motherhood, is education. In particular, there is a
link between higher education and delayed motherhood. From their analysis of
the women born in the mid 1950s in the Longitudinal Study (LS) of England and
Wales, Rendall and Smallwood (2003, p. 25) conclude that women “who obtained
a higher-education qualification began motherhood, on average five years later than
45
0.8 0.6 0.4 0.2
40
35
Mean age
30
25
20
Fig. 4.3 Age by which
successive quintiles of the 15
1925
1930
1935
1940
1945
1950
1955
1960
1965
1970
1975
1980
1985
whole cohort had entered

motherhood
(Source: ONS, 2005) Year
did women who did not obtain a higher qualification”. They also show that the pro-
portion of first-time mothers who have a higher qualification is greater at later ages.
As part of this project, Jenkins et al. (2008a) consider the timing of mother-
hood for women born in 1958, exploiting the richness of the fertility and employ-
ment history data available in the NCDS. As these data are longitudinal and the
participants in the survey have been followed for over 45 years it is possible to
consider the entry to motherhood for the cohort members themselves in terms
of information across their life course to date. Using event history analysis, i.e.
looking at the probability of having a birth of a given order, at a given time,
for women who had not (yet) had such a birth; Jenkins et al. (2008a) find the
strongest predictor of transition to a first birth is the woman’s education. The
chance of having a first birth is strongly negatively associated with the level
of education, even after controlling for a number of childhood background fac-
tors and making allowances for an otherwise unobserved propensity to progress
to the next birth. The next birth is defined as births from the second up to the
fourth.
Jenkins et al. (2008b) extended their work on the NCDS to consider the timing
of motherhood for members of the BCS70 cohort, taking birth orders up to the
third birth and allowing for the possibility of an effect of the aggregate labour
market on birth decisions. The association of early childbearing and individuals
with particularly poor labour market prospects (noted above) led to the hypothesis
that the incentive to avoid early motherhood might be stronger when labour market
prospects were better. The NCDS cohort encountered particularly adverse labour
markets when they were in their mid-twenties and mid thirties. For BCS70 mem-
bers these recessions occurred when they were in their mid teens and early twenties.
For the BCS sample, education remained the dominant explanatory variable for the
transition to first birth, but the effect was weaker than for the NCDS sample. For the
earlier cohort, women with tertiary education were, all else equal, seven times less
likely to embark on motherhood at a given time (if they had not already done so)
while the comparable estimate for women born in 1970 was 3.6, relative to those
with no qualifications.
The aggregate unemployment rate showed a significant relationship with the
hazard of the first birth only for the BCS70 cohort. The estimate was found to be
negative, contrary to expectation, and to earlier results, which pooled both cohorts’
birth histories up to 2000 only (Dex et al., 2005). According to the 2008 results,
‘bad times’ tend to discourage rather than bring births forward. Further investiga-
tion could be made of the possibility that unemployment effects vary within the
cohort, or that there were changes in the economic and policy environment in the
years after 2000 that were relevant.
Using data on both men and women in the NCDS and the BCS70, Kneale and
Joshi (2008) consider the patterns of postponement and the childlessness across
the two cohorts. Once again a major factor in both postponement and childless-
ness for women, and also men, was education, with the more educated becom-
ing parents later, if at all, than their less educated peers. However, they suggest
that if the 1970 cohort do turn out to have a higher incidence of childlessness
66 K. Hansen et al.
(as predicted), this may be due to the unintended consequences of delaying the
decision rather than an active decision not to have children, given the preferences
for parenthood expressed by those still childless at 34.
The MCS consists of data on around 19,000 children born across the UK in
2000/2001 (Hansen, 2008). Hawkes (2008) and Jayaweera et al. (2005) use data on
the mothers of the cohort members to consider possible determinants of the age at
first birth. The results of these regressions reflect the patterns in the earlier NCDS
and BCS70 cohorts, found by Jenkins et al. (2008a, b), among others, but unlike
the event history approach, the analysis is necessarily confined to women who have
actually become mothers. Hawkes (2008) finds that the age at first child is associated
with both antecedent and current disadvantage. Once again, the woman’s education
is a strong positive correlate of the age at first birth. In addition to low education,
those who enter motherhood earlier are more likely to have experienced disadvan-
tage in their own childhood, the separation of their own parents, spent time in care
and having an unemployed father at the age of fourteen.
Certain ethnic minority groups (especially Pakistani and Bangladeshi) are asso-
ciated with early motherhood. Being born outside the UK is surprisingly found to be
associated with later motherhood for some of this sample (though not to Pakistanis
and Bangladeshis). One possible explanation for this positive association of age at
motherhood and immigration is that the migration process itself has disrupted and
delayed family formation. Another is that migration is selective of more enterpris-
ing and career-motivated individuals, which may be underlined by the selectivity of
immigration controls. Finally, higher unemployment, measured locally and nation-
ally at the time of conception, is found to be associated with earlier motherhood.
This finding reflects those of Dex et al. (2005) on births up to 2000, but not in the
updated data set analysed by Jenkins et al. (2008b).
In summary, we confirm that young motherhood is strongly linked to indicators
of childhood disadvantage, such as having an unemployed father and experiencing
family break-up, as well as being associated with the situation in the labour market.
However, education plays a key role in the timing of motherhood with those who
invest in more education delaying their entry to motherhood, not only to extend their
enrolment in post compulsory education, but to pursue careers thereafter. It is likely
that maternal education will also be a key determinant in the labour market attach-
ment of a mother, and that the anticipation of a career after childbearing, signalled
by the level of qualifications, may play a role in decisions about when childbearing
starts. Therefore in the next section we turn to considering maternal employment.
Maternal Employment
Maternal employment, even in the early years of a child’s life, is now more usual
than not in the UK. Figure 4.4 shows the proportion of mothers working by
partnership status (Duffield, 2002). As shown in Fig. 4.4 by the age of the child,
maternal employment appears to be linked to partnership status. Across all of the
age groups those mothers who are single are less likely to be in work. This may
Fig. 4.4 Employment rates 90
Percentage of mothers
Couple Single
for partnered and single 80
women by age of youngest 70
dependent child, UK, spring 60
2002, not seasonally adjusted 50
(Source: Duffield, 2002) 40
30
20
10
0
<5 5–10 11–15 16–18 All with
dependent
children
Age of youngest dependent child
reflect the relative ease of combining employment with parenthood when there is a
second parent present, but it may not be a pure effect of partnership status alone. It
is likely to be confounded with the woman’s age, since early childbearing is associ-
ated with relatively unstable, if any, partnership, as well as low earning power (see
Fig. 4.2 and Hawkes et al., 2004). Thus, children who have a young mother are more
likely to have a mother with neither job nor partner. This situation will be linked to
a range of disadvantageous experiences for children including a very high risk of
family poverty and a reliance on means-tested benefits which can form a barrier to
labour market entry.
The Birth Cohort Studies can offer a longitudinal perspective on the secular
increase in the employment of mothers of young children. Table 4.2 shows the per-
centage of the members of the three Birth Cohort Studies who had an employed
mother at some point in the first five years of their life. Although the numbers are
not exactly comparable, they appear to have doubled from 29% in 1958–1963 to
around 58% in 2000–6, with most of the increase occurring before 1970–1975.
Table 4.2 Percentage of mothers in work during time child aged under five years, across three
birth cohorts
NCDS BCS MCS
Employment in first 5 years 29% 48% 58%

Base sample of mothers 13,966 11,474 14,396
Coverage GB UK
Source: Authors’ analysis.

NCDS reports mothers having any paid work before the child went to school. BCS70
reports mothers having any paid work at up to the age 5 interview who held a job which
started more than a year earlier. MCS reports current employment at age 5 interview.
It will include a few who only started after the 5th birthday, but is does not include
those who had worked before age five who were not currently employed. Percentage is
weighted.
Source: NCDS and BCS rates from Hansen et al. (2006), MCS numbers from Dex and
Ward (2008).
68 K. Hansen et al.
Cumulative % of whole sample

Fig. 4.5 Month of 0.6
return/entry to employment
after birth 0.5
(Source: Authors’ analysis)
0.4
0.3
0.2
0.1
0
<1 1 2 3 4 5 6 7 8 9
Month
Figure 4.5 shows the proportion of mothers in the MCS who had entered employ-
ment in the nine months after the birth of the cohort member. The steeper gradient
between three and four months matches with the end of the paid maternity leave at
the time of the study.
Table 4.3 presents some information on the hours of employment the MCS
mothers were undertaking in the first nine months of life. The majority were
employed in part-time jobs with very few undertaking very long hours.
Crosby and Hawkes (2007) consider the determinants of maternal employment
in the first nine months. The main determinants are maternal employment in the year
prior to the birth and maternal education. Those employed the year before the birth
are much more likely to be employed by nine months after the birth than those who
were previously out of the labour market. Of course, being in the labour market a
year before the birth is also likely to be related to maternal education. Maternal edu-
cation is also directly linked to early employment post birth, with the more educated
being more likely to enter employment in the first nine months. Other significant
predictors of non-employment include having no partner, younger (currently under
20) and older (currently over 35 or 40) or belonging to an Asian ethnic group.
In this strand of our work we have also made international comparisons with
MCS. This research looks for determinants of the timing of employment after
childbirth and attempts to infer the role of government policy in these decisions.
The results suggest that maternal employment, especially in the early years of
Table 4.3 Work schedule for mothers employed when child aged nine months, MCS
Hours Regularity
Part-time 72%
Full-time ( 31–39 hours) 23%
40+ hours per week 5%
Works regular, daytime hours 54%
Other pattern 46%
Unweighted n 7,954
Percentages are weighted

Source: Crosby and Hawkes (2007)
a child’s life, can also be linked to the structure of public policies. Crosby and
Hawkes (2007) examine differences in the UK (MCS) and the US (Early Child-
hood Longitudinal Study of Births (ECLS-B)) while Baxter et al. (2006) compare
the UK (MCS) with Australia (Longitudinal Study of Australia Children (LSAC)).
While womens’ education and earning power are associated with the chances of
getting back into the labour market by the end of the first year in all three coun-
tries, there are striking contrasts in patterns of entry and re-entry during those early
months.
Both international comparisons provide evidence of the role of the benefit system
and maternity leave policies in the timing of employment after child birth. The
nature of means-tested benefits appears to discourage employment for the poten-
tially low paid in the UK but not the US. In the UK the drive to reduce child poverty
may have had unintended consequences on the incentives of low paid mothers with
regard to employment. Maternity leave policies, applying to at least some of those
who were employed before the birth, appear to encourage returning to employment
after a birth, by permitting a delay, which is longer in the UK than US. In Australia,
many mothers stayed away from their job for most of the 12 month duration of
their unpaid leave of absence, whereas in the UK, those with maternity leave rights
tended to go back to work either at the expiry of paid leave after four months, or after
unpaid job protection ended at seven months. In the US, where mothers experience
a lack of paid maternity leave and a lack of state benefits, most women entered
employment very soon after birth. Differentials by education are exacerbated in that
it sorts women into either low paid jobs or jobs with better than statutory maternity
leave pay. This is the case in the UK.
Other important determinants of maternal employment in the first nine months of
the child’s life include: being the first born, being a twin/triplet, and low birth weight
which all tend to postpone mothers’ employment. Lone motherhood predicts later
entry into post-birth employment for UK (but not US) mothers.
In summary, we find that the maternal employment is strongly linked to the
mother’s characteristics especially her education. International comparison between
the UK and the US/Australia suggest that employment after child birth is also con-
ditioned by the benefit system and maternity leave policies.
Childcare and Child Outcomes

Once a mother has decided to enter employment after childbirth one of the most
important decisions she has to make is who shall look after her child whilst she is at
work. Table 4.4 presents evidence of the childcare use in the early years for mothers
across the three cohort studies who worked when the child was under 5. This shows
the growing importance of formal childcare in the pre-school years across the three
cohorts.
Table 4.5 presents results from a National Survey of Childcare Use (Bryson et al.,
2006). For those under five years much of the childcare is informal. Of this informal
provision, grandparents appear to be undertaking the lion’s share of the task.
70 K. Hansen et al.
Table 4.4 Any source of childcare for working mothers of young children across the three birth
cohorts
NCDS BCS70 MCS∗∗

1958–1963 1970–1975 2000–2005
(0–4 years) (0–4 years) (0–5 years)
Informal only 73.5 19.9 14.3

Formal only 39.0∗ 17.7
Both formal and informal 41.1∗ 63.9
At least some formal 26.5 81.6
Observations 4,144 4,246 12,392
Sources: NCDS and BCS numbers from Hansen et al. (2006); MCS numbers from Hansen et al.
(2008, Table 5.5).
Informal care includes self provision and care provided by partners, grandparents, other relatives
and friends. Formal care includes nanny, childminder, nursery, crèche provision. Missing numbers
in the columns for NCDS and BCS70 represent the different ways of collecting the data due to the
growing interest in child care issues.
∗
Formal only in BCS70 and the mixed category each reduces to 20% if playgroup only is dis-
counted (see Table 4.7b in Hansen et al., 2006).
∗∗
The MCS figures are based on all families present in all three sweeps of the MCS, not just those
with working mothers. They also include nursery school as a form of formal childcare.
Table 4.5 Use of provider types in the last week, by age of child
0–2% 3–4% 5–7%
Early years provision and formal childcare

Nursery school 2 10 +
Nursery class 1 22 +
Reception class 0 28 5
Day nursery 18 12 +
Playgroup or pre-school 9 18 +
Childminder 5 5 4
Nanny or au pair 1 2 1
Babysitter 3 2 3
Out-of-school club on-site + 2 13
Out-of-school club off-site + 1 4
Informal childcare
My ex-husband/wife/partner 3 4 5
The child’s grandparent(s) 29 26 22
The child’s older brother/sister 1 1 2
Another relative 7 6 5
A friend or neighbour 5 6 9
Base: All families Umveighted base 1451 1507 1348
Source: Bryson et al. (2006).
The Department for Children, Schools and the Family (DCSF), formerly the
Department for Education and Skills (DfES), publishes data on the use of pre-school
provisions for those who are three or four years old. Figure 4.6 presents these data
graphically, showing the majority of those aged three to four in England and Wales
Fig. 4.6 Percentage of 100

children aged three and four
taking up early education
Percentage of children
places by type of early 80
education provider
(Source: Department for
Education and Skills, 2006) 60
40
20
0
2002 2003 2004 2005 2006
Year
Maintained nursery and primary schools
Independent schools
Private and voluntary providers
attend a state nursery or primary school (nearly 60% in all years between 2002 and
2006), with a minority attending independent schools (5% a year) and the remainder
attending other private and voluntary places.
Why are these patterns important? We are concerned with the impact of these
three decisions (timing of motherhood, maternal employment and childcare choice)
on the outcomes for children (cognitive, behavioural and health). This will provide
us with an indication of possible externalities for children of the government policies
towards parents. Externalities could be a positive or negative in terms of outcomes
for children. We present here the results of our project which consider the impact
on child outcomes (cognitive, behavioural and health) on the timing of motherhood,
maternal employment and childcare choices.
Firstly, Verropoulou and Joshi (2008) use the NCDS second generation data
to consider the association between pre-school employment for the NCDS cohort
members on the later outcomes for their children, observed in 1991. They find that
reading is slightly poorer where less educated mothers work in the child’s first year
of life. They find few other interactions with employment, but do detect intergener-
ational transmission of behavioural as well as cognitive characteristics. The age at
first birth has a significant independent association only with one of the four out-
comes investigated – the child’s maths score which is lower for children with moth-
ers aged 20 or less. As all of the children in this study are the offspring of relatively
young mothers, it may be difficult to generalize from these findings. However, in so
far as the associations of mothers’ employment with indicators of child development
are mixed and minor, they are broadly similar to those estimated for the children of
mothers of all ages in the ALSPAC study (Gregg et al., 2005).
72 K. Hansen et al.
Secondly, Crosby and Hawkes (2008) considered the association between

employment in the first nine months of life and child cognitive and behavioural out-
comes measured at age three years old. The results using the MCS show that mater-
nal employment has a positive association with child cognitive outcomes (measured
using a vocabulary test and a school readiness measure) and child behaviour. How-
ever, it is negatively associated with child health, as measured by the body mass
index (BMI). After controlling for the selection process into employment, only the
association between child health and maternal employment is maintained, albeit
weakened. The results suggest that maternal education rather than maternal employ-
ment in the UK is the main driver for all four child outcomes. Of course education
also plays a strong role in the selection process into employment.
In our third study, Hansen and Hawkes (2009) consider the role childcare choice
has on child cognitive and behavioural outcomes using the MCS. This paper finds
that for those in employment, different types of childcare have different impacts on
child cognitive and behavioural outcomes at age three. Formal childcare is found
to have positive associations with school readiness measures compared to other
types of childcare. This is true particularly for children from disadvantaged back-
grounds. Care by grandparents appears to have positive effects on a child’s vocab-
ulary development but is also negatively associated with behaviour, largely through
poorer peer relations.
Finally, we examine mother’s age at first birth; whether and when she took up
employment in the child’s first nine months; what main childcare arrangements were
made at nine months; and measures of child development at age three. These are
shown in Table 4.6, for mothers of first-born children in the MCS. Although we
have estimated age at first motherhood for cohort children who were not first-born,
we do not have such detailed data on employment and childcare in the first year of
their eldest sibling’s life. The three indicators of child development are the Bracken
School Readiness, British Ability Scales (BAS) Naming Vocabulary, and the Total
Difficulties Score of the Strengths and Difficulties Instrument. All indicators appear
in terms of standard deviations, and are adjusted for age at interview by inclusion
of child’s age in the battery of controls noted in Table 4.6. Bear in mind that the
Difficulties Score falls when child behaviour improves.
Taken singly, each of these three factors was significantly associated with better
child outcomes: older mothers, employment in first nine months and use of formal
childcare. However, the multiple regression model enables us to see how far these
positive associations are attributable to other things. The mother’s level of quali-
fication is a dominant predictor of all the outcomes. Putting all the factors in the
model together moderates the differences attributable to age at first motherhood and
mode of childcare, and completely accounts for any differences by time of return
to employment in the first year. Once we allow for other factors, particularly for
the education of their mothers, there are few significant differences between chil-
dren whose mothers were employed in their first nine months and those who were
not. Children whose mothers worked and had been in formal group care had higher
school readiness scores and lower vocabulary scores than other children. They also
had significantly better behaviour ratings than children whose mothers stayed at
4
Table 4.6 Child development at age three analysed by mother’s age, employment and childcare in the first year, and other controls: multiple regression of
mothers whose cohort child was their first
School Readiness (Bracken) Naming Vocabulary (BAS) Total Difficuties Score (SDQ)
Coefficient Standard error Coefficient Standard error Coefficient Standard error
Mother’s qualifications (baseline: no

qualifications)
0.20 0.07 ∗∗∗ 0.19 0.08 ∗∗ −0.17 0.09 ∗
NVQ Level 1
0.28 0.06 ∗∗∗ 0.27 0.07 ∗∗∗ −0.32 0.07 ∗∗∗
NVQ Level 2
0.42 0.06 ∗∗∗ 0.31 0.06 ∗∗∗ −0.45 0.07 ∗∗∗
NVQ Level 3
0.54 0.06 ∗∗∗ 0.42 0.07 ∗∗∗ −0.52 0.07 ∗∗∗
NVQ Level 4
0.68 0.10 ∗∗∗ 0.39 0.09 ∗∗∗ −0.46 0.09 ∗∗∗
NVQ Level 5
0.34 0.15 ∗∗ 0.03 0.13 −0.31 0.15 ∗∗
Overseas and other qualifc’s
Mother’s age at ( first) birth (baseline:
14−19)
0.14 0.06 ∗∗ 0.06 0.05 −0.06 0.06
20−24
0.30 0.06 ∗∗∗ 0.23 0.05 ∗∗∗ −0.22 0.06 ∗∗∗
25−29
The Timing of Motherhood, Mother’s Employment and Child Outcomes
0.41 0.06 ∗∗∗ 0.30 0.06 ∗∗∗ −0.22 0.06 ∗∗∗

30−34
0.35 0.08 ∗∗∗ 0.28 0.06 ∗∗∗ −0.26 0.07 ∗∗∗
35−39
0.50 0.22 ∗∗ 0.24 0.15 −0.12 0.15
40+
73
Table 4.6 (continued)
74
School Readiness (Bracken) Naming Vocabulary (BAS) Total Difficuties Score (SDQ)
Coefficient Standard error Coefficient Standard error Coefficient Standard error
Returned to employment (baseline: not

returned to employment during the
first nine months)
−0.18 0.11 0.08 0.09 0.20 0.10 ∗
First three months
−0.19 0.11 ∗ 0.10 0.08 0.17 0.09 ∗
Between four and six months
−0.12 0.11 0.12 0.09 0.15 0.10
From seven months
Child care whilst mother employed

(baseline: not using child care)
0.02 0.12 −0.22 0.11 ∗ −0.30 0.12 ∗∗
Informal non group
0.00 0.11 −0.07 0.09 −0.23 0.11 ∗∗
Father/partner
0.13 0.11 −0.03 0.08 −0.15 0.10
Grandparent
0.08 0.12 −0.15 0.09 −0.24 0.10 ∗∗
Formal non group
0.23 0.11 ∗∗ −0.18 0.09 ∗∗ −0.29 0.10 ∗∗∗
Formal group
R squared 0.23 0.19 0.19
Observations 4,647 4,870 5,101
Source: Authors’ analysis.

All regressions control for the following: Developmental controls: Age at assessment, female, low birth weight, multiple birth; Family Controls: Ethnicity,
English only spoken at home, partner, employed partner, household income, owner occupied housing, mother s depression at nine months, mother employed
full time before nine months, region. Only first born children used in this analysis. ∗∗∗ 1% ∗∗ 5% ∗ 10%
K. Hansen et al.
home, as did those in most other types of childcare except grandparent. Looking at
age at motherhood, child outcomes are worst for those who were teenagers at birth
(the reference category) and improved as the mothers age rises to some point in the
thirties. The improvements reverse somewhat for the small number of mothers hav-
ing their first child in their forties. The two cognitive scores peak among mothers
aged 30 to 34, and the best behavioural adjustment, other things equal, found among
children whose mothers were aged 35 to 39. After these ages the numbers having
first births tails off (compare Fig. 4.3). The strength of the association with age at
motherhood can be compared with that of education.
For the school readiness score, the peak difference between having a teenage
mother and one whose first birth was at age 30 to 34 (0.41) is equivalent to the
difference between a mother having no qualifications and A levels (NVQ 3). Here
a degree is ‘worth’ a premium over no qualifications of 0.54 and a post-graduate
qualification – NVQ5 – a premium of 0.68. The child’s vocabulary score is not
so strongly related to either education or maternal age, but their 0.30 differential
between mothers in the early thirties and teenagers corresponds to the premium for
NVQ 3. Among age groups of mothers, those aged 35–39 had children with the best
behaviour scores, 0.26 below the behaviour problems of the children of teenagers.
This can be compared to the indicator with the biggest differential, mothers’ educa-
tion. The behaviour scores for children of mothers with NVQ 4 or 5 was 0.52 below
parents with no qualifications. Comparing a mother who had both tertiary qualifica-
tions and later entry to motherhood (as most do) with a mother who had her child
as a teenager and had no qualifications would involve adding the qualifications and
age terms. According to this model therefore, a larger part of the developmental
lead observed in the children of older mothers is attributable to their mothers’ edu-
cation, but there are more disadvantages to having a very young mother than those
associated with her lack of education, or correspondingly, more advantages associ-
ated with later motherhood than just those attaching to maternal education. These
advantages may be other material, social or psychological ‘capital’ not adequately
captured in the model, but they do suggest there could be a gain for the child if its
arrival is delayed until at least well into the twenties.
Conclusions
In this final section, we respond to the three questions asked at the outset:
What are the Main Determinants of the Age at Motherhood,

and Have these Changed Between the Cohorts?
Later motherhood in Britain has involved a widening of the span of ages at which
women have their first child. This has gone along with widening educational and
employment opportunities for women and increasing childbearing outside marriage.
76 K. Hansen et al.
Education drives fertility beyond those years of post-compulsory enrolment

which lead to qualifications but inhibit early fertility. Whether education is enlight-
ening or merely a way of signalling better access to the labour market, it can be
argued that, by raising the opportunity costs of motherhood it creates an incentive
to defer beyond years of study, if not avoid it. The case for the widening in ages
being driven by dispersion in educational achievement rather than vice versa can be
made by arguing that women with more human capital, or aspirations to accumulate
it, have a greater incentive to defer motherhood than those with poorer educational
or career prospects. Much early motherhood may be unintended, but since fertility
control, including abortion, has become available to those with the motivation and
knowledge to use it, it will tend to be those who have least to lose who will become
mothers soonest.
The modelling of the hazard of having a first birth in both the 1958 and 1970
cohorts found that a woman’s education was a strong predictor of entering moth-
erhood even after allowing for measured (and unmeasured) characteristics and
abilities of the women with which it is correlated. However, the strength of this
predictor was weaker in the second cohort, where more women had tertiary educa-
tion. In the analysis of the Millennium cohort, whose mothers come from a number
of birth cohorts themselves, we also found poor education and childhood disad-
vantage to be predictors of early motherhood, as too was ethnic variation. The
NCDS and MCS studies reach opposite conclusions in the attempt to associate poor
labour market conditions with early births, but perhaps the structure of constraints is
changing.
What are the Main Determinants of Maternal Employment

in a Child’s Early Life?
Here again we find the woman’s education strongly associated with her employ-
ment career, both before and in the year after the Millennium cohort child’s birth.
Education predicts independently employment in those early months of life and
employment before the birth of the child. Higher employment and higher labour
force attachment are bound up with older motherhood. Over and above the asso-
ciation with employment, education and partnership status, the modelling detected
some further minor differentials by (current) age of mother. International compar-
isons helped to reveal how mothers’ employment behaviour in the UK is structured
by state policies. Maternity leave provisions structure the timing of mothers’ return
to work, and the provisions of the means-tested benefits presumably account for the
lower rates of employment of lone mothers in the UK than the US, and than part-
nered mothers in the UK. An important consideration in the maintaining of career
continuity after a birth is the arrangements a mother is able to make for childcare.
The higher purchasing power of those with high wages and higher qualifications is
crucial in this respect.
What are the Impacts of the Childcare Decisions, Maternal

Employment and Age at Motherhood on Cognitive and
Behavioural Outcomes for Children?
Until the National Childcare Strategy inaugurated in 1998, most working mothers
in the UK used informal childcare before their children reached school age. The
mothers of the Millennium Cohort encountered more group nurseries and a greatly
expanded nursery education provision once their children reached the age of three.
During their child’s first year, however, relatively few used formal group day care,
and grandparents were an important source of care. Analysis of child outcomes for
children of the 1958 cohort, assessed in 1991, suggests little impact on the develop-
ment of those children in that context. The investigation of evidence emerging on the
development of children born in the Millennium, and assessed at age three (mostly
in 2004), also finds little direct impact of early maternal employment on child out-
comes, although there are signs of some types of child care being associated with
better results on some outcomes. The pattern is nuanced in both the study of the
children of NCDS members and the results presented here on MCS, it is the verbal
score which has a significant negative association with early employment. The latter
does not apply if the care arrangement is informal, particularly with a grandparent.
It is too early to see how the experience of group childcare and nursery education at
ages 1 to 4 may be traced in child outcomes. While there seems to be little gained or
lost in terms of child development if the mother maintains career continuity, there
does seem to be a considerable advantage to the child in being born to a mother
with a reasonable level of education as well as one who is closer to age 30 than age
20. This confirms the supposition that the intergenerational transmission of social
advantages is playing out in the most recent cohort to be studied, as well as its
predecessors.
Policy Implications
This research has shown that education is a key correlate of the life course of both
women and their children. Whilst maternal age at first birth, maternal employment
and childcare play a role, maternal education appears to be a very strong deter-
minant of future success. To what extent it is a fundamental determinant, serving
to enlighten and raise skills in production and childrearing; a reflection of capa-
bilities inherited from home rather than school; or a signal for social sorting, has
not been fully established. Whatever lies behind educational attainment it accounts
for much of the difference in outcomes between early and late mothers and in the
development of their children. This also suggests that policies which are successful
in limiting educational failure, such as provision of basic skills and prolonging and
supporting years in education are likely to stem the polarization between early and
late mothers and their children, and have the positive externality of helping the next
generation.
78 K. Hansen et al.
The independent association of young age at motherhood with poorer or delayed

cognitive and behavioural scores at age three may just reflect unmeasured disad-
vantaged antecedents of the minority of women who become mothers in their teens
and early twenties. However, the very fact of their youth, perhaps through a lack of
maturity and of unmeasured current economic and social resources, may compound
the disadvantages facing their children. If so, government policies to reduce unin-
tended teenage motherhood should, if successful, improve the prospects for both the
women and children involved.
To the extent that early motherhood has its own costs, avoiding it would be an
advisable policy for the private individual as well as public intervention, which can,
after all, only inform and support private decisions. There is no evidence here to sug-
gest that the gains to postponing motherhood continue indefinitely, at least beyond
the early thirties.
We find little evidence for maternal employment in the first year systematically
either harming or promoting child development. We find formal childcare having
both positive and negative associations with different aspects of child development.
This suggests that policies making childcare available to mothers across the spec-
trum are likely to be beneficial to both mothers and children, but that attention needs
to be paid to its quality. The development of integrated childcare centres under the
Sure Start programme and other childcare policies aimed at improving the quality of
childcare received may help have positive spillover effects on some child outcomes
and contribute to reversing the polarity between advantaged and disadvantaged fam-
ily backgrounds.
References
Allen, I. and Bourke Dowling, S. (1998) Teenage Mothers: Decisions and Outcomes, Policy Stud-
ies Institute, London.
Barrett, G. and Wellings, K. (2002) What is a ‘planned’ pregnancy? Empirical data from a British
study, Social Science & Medicine, 55(4): 545–557.
Baxter, J., Joshi, H., Sanson, A. and Ward, K. (2006) Mother’s employment and mental health in
the first year of a child’s life: A contrast between two maternity leave regimes, Presentation at
the International Conference on Child Cohorts, Oxford.
Blanden, J., Goodman, A., Gregg, P. and Machin, S. (2005) Changes in intergenerational mobility
in Britain, in Corak, M. (ed.) Generational Income Mobility in North America and Europe,
Cambridge University Press, Cambridge, pp. 112–147.
Bryson, C., Kazimirski, A. and Southwood, H. (2006) Childcare and Early Years Provision:
A Study of Parents’ Use, View and Experience, DfES Research Report No 723, National Centre
for Social Research, London.
Buxton, J., Clarke, L. Grundy, E. and Marshall, C.E. (2005) The long shadow of childhood: associ-
ations between parental social class and own social class, educational attainment and timing of
first birth; results from the ONS Longitudinal Study, Population Trends, 121(October): 17–26.
Bynner, J. and Joshi, H. (2007) Building the evidence base from longitudinal data: the aims, content
and achievements of the British Birth Cohort Studies, The European Journal of Social Science
Research, 20(2): 159–179.
Bynner, J. and Parsons, S. (1999) Young People not in Employment, Education or Training and
Social Exclusion: Analysis of the 1970 Cohort Study, Social Exclusion Unit, London.
Chandola, T., Coleman, D.A. and Hiorns R.W. (2002) Distinctive features of age-specific fertility
profiles in the English-speaking world: common patterns in Australia, Canada, New Zealand
and the United States, 1970–98, Population Studies, 53: 317–329.
Crosby, D. and Hawkes, D. (2007) Cross-national research using contemporary birth cohort stud-
ies: a look at early maternal employment in the United Kingdom and United States, Interna-
tional Journal of Social Research Methodology, Special Issue - Child Cohort Studies, 10(5):
379–404.
Crosby, D. and Hawkes, D. (2009) Child outcomes and early maternal employment: evidence for
the UK and the US, submitted to Child Development.
Department for Education and Skills (2006) Provision for Children Under Five Years of
Age in England: January 2006 (Final), This Statistical First Release updates and extends
the information provided in SFR 17/2006 Provision for Children Under Five Years of
Age in England: www.dfes.gov.uk/rsgateway/DB/SFR/s000674/release.shtml is contained in
www.dfes.gov.uk/rsgateway/DB/SFR/s000674/index.shtml.
Dex, S. and Ward, K. (2008) Parents Employment and Education in Hansen and Joshi (eds) MCS3
User Guide to Initial Findings, Centre for Longitudinal Studies, London.
Dex, S., Gustafsson, S., Jenkins, A., Joshi, H., Kenjoh, E., Killingsworth, M.R. (2005) Mothers
changing labour supply in Britain, the US, and Sweden, in Goméz-Salvador, R. Lamo, A.
Petrongolo, B. Ward M. and Wasmer E. (eds.) Labour Supply and Incentives to Work in Europe,
Edward Elgar, Cheltenham, UK, pp. 115–150.
Duffield, M. (2002) Trends on Female Employment 2002, Labour Market Trends, 110 (November):
605–616.
Ermisch, J., and D. Pevalin. (2003a) Who has a child as a teenager? ISER Working Paper 2003–30,
Institute of Social and Economic Research, University of Essex, Colchester.
Ermisch, J., and Pevalin, D. (2003b) Does a teen birth have longer-term impact on the mother?
Evidence for the 1970 British Cohort Study, ISER Working Paper 2003–28, Institute of Social
and Economic Research, University of Essex, Colchester.
Ermisch, J.F. and Pevalin D.J. (2005) Early motherhood and later partnerships, Journal of Popula-
tion Economics, 18: 469–489.
Feinstein, L. (2003) Inequality in the early cognitive development of British children in the 1970
Cohort, Economica, 70 (277): 73–97.
Geronimus, A. and Korenman, S. (1992) The socioeconomic consequences of teen childbearing
reconsidered, Quarterly Journal of Economics, 107: 1187–1214.
Gregg, P., Harkness, S. and Machin, S. (1999) Child Development and Family Income, Joseph
Rowntree Foundation, York.
Gregg, P., Washbrook, E., Propper, C. and Burgess, S. (2005) The effects of a mother’s
return to work decision on child development in the UK, The Economic Journal, 115(501):
F49–F80.
Hansen, K. (2008) Millennium Cohort Study. A Guide to the Datasets, Centre for Longitudinal
Studies, Institute of Education, University of London, London.
Hansen, K. and Hawkes, D. (2009) Early childcare and child development, Journal of Social Policy,
38(2): 211–239.
Hansen, K., Joshi, H. and Verropoulou, G. (2006) Childcare and mothers’ employment: approach-
ing the Millennium, National Institute Economic Review, 195(January): 84–102.
Hawkes, D. (2008) The UK Millennium Cohort Study: the circumstances of early motherhood, in
Edwards, R. (ed.) Researching Families and Communities: Social and Generational Change,
Routledge, Abingdon, pp. 147–163.
Hawkes, D. (2008) The socio-economic consequences of early child bearing: evidence from a
sample of UK female twins, submitted to Journal of Population Economics.
Hawkes, D., Joshi, H. and Ward, K. (2004) Unequal entry to motherhood and unequal starts in life:
evidence from the first survey of the UK Millennium Cohort, Working Paper 6, CLS Cohort
Studies, London, www.cls.ioe.ac.uk/core/documents/download.asp?id=299&log˙stat=1
Hobcraft, J., and Kiernan, K.E. (2001) Childhood poverty, early motherhood and adult social exclu-
sion, The British Journal of Sociology, 52(3): 495–517.
80 K. Hansen et al.
Jayaweera, H., Joshi, H., McFarlane, A., Hawkes, D. and Butler, N. (2005) Pregnancy and child
birth, in Dex, S. and Joshi, H. (eds.) Children of the 21st Century from Birth to Nine Months.
The UK Millennium Cohort Study Series, The Policy Press, Bristol, pp. 109–132.
Jenkins, A., Joshi, H., and Killingsworth, M. (2008a) The relationship between educational attain-
ment and the timing of first and higher-order births. An Event History Analysis of the National
Child Development Study, Unpublished paper presented at European Society of Population
Economics (ESPE) conference, June, www.econ.ucl.ac.uk/espe2008/
Jenkins, A., Joshi, H., and Killingsworth, M. (2008b) Educational attainment, labour mar-
ket conditions and the timing of first and higher-order births in Britain, Unpublished
Paper Presented at European Association of Labour Economists (EALE) Conference, Sept.
www.eale.nl/Conference2008/Programme/PapersC/add72162˙tqMZ5V8zNr.pdf
Kiernan, K.E. (1992) The impact of family disruption in childhood on transitions made in young
adult life, Population Studies, 46: 213–234.
Kiernan, K.E. (1997) Becoming a young parent: a longitudinal study of associated factors, British
Journal of Sociology, 48: 406–428.
Kiernan. K.E. and Diamond, I. (1983) The age at which childbearing starts -a longitudinal study,
Population Studies, 37(3): 363–380.
Kneale, D. (2009) Early parenthood: definition and prediction in two British cohorts, in Stillwell,
J., Coast, E. and Kneale, D. (eds.) Fertility, Living Arrangements, Care and Mobility: Under-
standing Population Trends and Processes Volume 1, Springer Dordrecht, pp. 81–103.
Kneale, D. and Joshi, H.E. (2008) Postponement and childlessness – evidence from two British
cohorts, submitted to Demographic Research.
Liao, T.F. (2003) Mental health, teenage motherhood, and age at first birth among British women
in the 1990s, ISER Working Paper 2003–33, Institute of Social and Economic Research,
University of Essex, Colchester.
Office for National Statistics (2005) Dataset PBH41A Taken from Table 1.6 Published in Birth
Statistics 1837–1983, Historical series FM1 No 13, and Vol. No’s 27, 28, 29 and 33.
Office for National Statistics (2007) Birth Statistics Series FM1 no. 35 Review of the Registrar
General on Births and Patterns of Family Building in England and Wales, 2006, Office for
National Statistics, Newport.
Pevalin, D.A. (2003) Outcomes in childhood and adulthood by mother’s age at birth: evidence from
the 1970 British Cohort Study, Working Paper 2003-31, ISER, University of Essex, Colchester.
Rendall, M.S., Couet, C., Lappegard, T., Robert-Bobee, I., Ronsen, M. and Smallwood, S. (2005)
First birth by age and education in Britain, France and Norway, Population Trends, 121: 27–34.
Rendall, M.S., Joshi, H., Lynch, K., Ekert-Jaffe, O. and Mougin, R. (2008) Universal versus eco-
nomically polarized change in age at first birth: a French-British comparison, Rand Corporation
Working Paper WR568, http://www.rand.org/pubs/working papers/WR568.
Rendall, M. and Smallwood, S. (2003) Higher qualifications, first-birth timing, and further child-
Schoon, I. (2006) Risk and Resilience: Adaptations to Changing Times, Cambridge University
Press, Cambridge.
Smallwood, S. (2002) New estimates of trends in births by birth order in England and Wales,
Population Trends, 108: 32–41.
Social Exclusion Unit (SEU) (1999) Teenage Pregnancy, TSO, London.
Sylva, K., Melhuish, E., Sammons, P., Siraj-Blatchford, I. and Taggart, B. (2004) The Effective
Provision of Pre-school Education (EPPE) project, Technical Paper 12 – The Final Report,
DfES/Institute of Education, University of London, London.
Van de Kaa, D. (1987) Europe’s second demographic transition, Population Bulletin, Washington
DC.
Verropoulou, G. and Joshi, H. (2008) Does mothers’ employment conflict with child develop-
ment? Multi-level analysis of British mothers born in 1958, submitted to Journal of Population
Economics.
Ward, K. and Dex, S. (2007) Parental employment, in Hansen, K. and Joshi, H. (eds.) The Millen-
nium Cohort Study Second Survey: A User’s Guide to Initial Findings, Centre for Longitudinal
Studies, London.
Chapter 5
Early Parenthood: Definition and Prediction
in Two British Cohorts
Dylan Kneale
Background
Since the baby boom of the 1960s, births in the United Kingdom (UK) have been
falling in number while age at first birth has risen. Low and late fertility is an impor-
tant part of the second demographic transition (Van de Kaa, 1987). In the first years
of the new millennium, while the total fertility rate has fluctuated slightly and has
once again begun to rise, or catch up, the tendency to postpone entry to motherhood
has persisted with a continued rise in average age at first birth (Office for National
Statistics, 2007). However, this rise has not occurred as a neutral shift towards older
ages. While teenage fertility rates have only dropped slightly since peaking in the
late 1990s, this pace of change has not matched the dramatic rise in fertility among
older age groups (Office for National Statistics, 2007). In 1996, the age-specific
birth rate for teenagers stood at 29.7 per 1,000 women, dropping to 26.6 in 2006.1
In the same period, the rate increased from 37.5 to 53.8 births per 1,000 women
aged 35–39 years. A picture of decreasing fertility in the twenties is painted when
we recognise that by the age of 30 years, 67% of women born in 1960 would have
become mothers while for women born just fifteen years later, in 1975, this estimate
had dropped by 10 to 57% (Office for National Statistics, 2007).
Such decreases are usually attributed to postponement, as opposed to large-scale
avoidance of parenthood (Kneale and Joshi, 2008). Put in succinct terms, this almost
denotes “a move towards older fertility for the majority and early fertility for the
minority” (Hadfield et al., 2007, p. 255). This unequal shift in the fertility sched-
ule is understood to be representative of social polarization in age at first birth
(Joshi, 2007, Kneale and Joshi, 2008). Older first-time mothers are associated with
a range of advantageous characteristics not shared by early first-time mothers. It is
D. Kneale (B)
e-mail: D.Kneale@ioe.ac.uk
1 It can be assumed that most teenage births would represent first births. To the author’s knowledge,
statistics on birth order for births outside marriage are not collected. We therefore demonstrate
these trends using statistics on all births.

82 D. Kneale
this polarization in the characteristics and outcomes of first-time mothers –and by

extension, fathers and their children – that is the cause of continued concern about
the transition into early parenthood.
One of the widespread conventions of the literature on early parenthood is that
parenthood stops being ‘early’ at the age of 20 years. There are various definitions
of ‘teenage’ motherhood in the literature – pregnancy under 16, 18 or 20 – as well as
a more literal cut-off of giving birth while still aged 19 or under (Social Exclusion
Unit, 1999; Birch, 1996). The concern here is why the line should be drawn at age
20. Despite the shift of the fertility schedule, albeit somewhat unbalanced, towards
older ages, this convenient definition of teenage parenthood remains a pervasive
concept in studies, with very few extending the cut-off point for early parenthood
beyond this.2 In studies examining teenage parenthood, there is little or no justifica-
tion given as to why the clock stops at 20 years.3 This is despite the fact that many
of our parents and grandparents may have been teenage first-time parents them-
selves (Geronimus, 1997). In fact, in her later work, Geronimus describes teenage
parenthood as more a political tool than a social construct (Geronimus, 2003), and
certainly, there are grounds to question this distinction. Not only does the age 20
threshold come at a convenient point in the coding of ages into five-year bands,
but the term ‘teenage’ has other connotations – of adolescence, immaturity and not
being old enough for the ‘adult’ role of a parent. A ‘teenage parent’ is almost a
contradiction in social terms.
In terms of outcomes, there are only a few studies where the distinction of
teenage and early parenthood has been explored. In their study, Hobcraft and
Kiernan found that the widest gulf in terms of adult outcomes was between those
having a first birth under 23 years and those between 23 and 32 years, although
this was reinforced by ‘teenage’ parenthood (Hobcraft and Kiernan, 2001). As men-
tioned however, studies that experiment with a definition of ‘early’ in terms of out-
comes are few and far between. Some researchers who have taken a counterfactual
approach to the timing of parenthood by asking ‘what would have happened had the
teenage (mother) delayed childbearing?’ have found no substantial benefit to delay-
ing motherhood among this group (Goodman et al., 2004) and have even found early
motherhood to be a beneficial strategy (Hotz et al., 2004). This again would appear
to suggest that joining the labels of teenage and parent in such an unchallenged
way is the product of a form of manufactured risk. Throughout this chapter, it is the
hypothesis that focus on teenage parents has stemmed from a stigmatised view of
the ‘correct path to parenthood’ (Hadfield et al., 2007). This focus has seemed dis-
proportionate given their low and diminishing prevalence. It is the proposition here
that in terms of predictors, their characteristics do not vary significantly from early
parents in their twenties.
2 The works of Robson and Berthoud (2006) and Hobcraft and Kiernan (2001) provide two exam-
ples of a more inclusive definition of ‘early’ parenthood.
3 This is addressed fully in Kneale (2008).
5 Early Parenthood: Definition and Prediction in Two British Cohorts 83
To examine the validity of the teenage construct, this research uses a number of
definitions of ‘early’ through which to examine transition to early parenthood as
well as using different ways of modelling ‘early’. Although this chapter is focused
firmly on whether age 20 is still a meaningful boundary as concerns the transition
to parenthood, it can also be seen as a contribution to the literature on the timing
of ‘youth transitions’ in the life course generally, with many other aspects of adult
roles, in the labour and housing markets for example, being ‘delayed’ well beyond
age 20 (Pollock, 2008).
Predicting Early Parenthood
The labels ‘early parent’ or ‘teenage parent’ apply to a diverse group of people.
Indeed, it has been speculated that the only commonality binding early parents is just
that – that they were young when having children (Harden et al., 2006). However,
there is a considerable body of quantitative and qualitative studies finding recurring
themes and patterns that have predictive power over the timing of parenthood. These
can be grouped under the following four headings4 and while these groups are not
exhaustive, they serve as a useful framework of the factors that are known to be
significant. This framework applies to becoming a parent. It is not within the range
of this study to examine factors associated with becoming pregnant but avoiding
parenthood (abortion) or to examine predictors of avoiding conception (abstinence
and contraception).
Educational Factors
Educational underachievement and cognitive ability have been linked in several

studies to an increased likelihood of becoming an early parent (Ermisch and Pevalin,
2003b; Harden et al., 2006; Jaffee et al., 2001). The pathway often suggested is one
of low qualifications leading to poorer labour market prospects. This in turn lowers
the opportunity costs of having children early (Ermisch and Pevalin, 2003b). This
pathway has applied to the case of early motherhood. For early fatherhood, educa-
tional factors are included as a matter of course, although there is little theoretical
development of such a causal pathway for men as in the case of women. In particu-
lar, while low educational achievement is likely to lead to poorer trajectories in the
labour market; there is no theorised substitution of parenthood for career as is the
case for women. A possible pathway that could form from the view that education
may be important is the partnership market. Low levels of education may, through
‘assortative mating’ (or homogamy), go along with a partner with low educational
level for whom the opportunity cost of early childbearing is low, or who is herself
4 References represent only a selected group of studies that examine these predictors.
84 D. Kneale
impatient to start a family. Alternatively, lower educational achievement may be

associated with a greater preponderance to risk taking behaviour which may include
fathering early pregnancies.
A limited number of studies have explored the effect of education further through
examining the effect of disliking school on early pregnancy (Bonell et al., 2005,
2007; Imamura et al., 2007; East et al., 2006). In Bonell’s 2005 study, dislike
of school was found to be a potent predictor of early teenage (under 16) preg-
nancy and replaced other personal factors. However, this association was statisti-
cally accounted for by the inclusion of socioeconomic features (Bonell et al., 2005).
In East and colleagues’ study, school orientation was measured as achievement and
ambition for higher education, both of which were insignificant alongside other fac-
tors (East et al., 2006). In qualitative research, links have also been made between
dislike of school and pregnancy, thought to operate along a pathway to reduced
opportunity costs in having children early (Arai, 2003). In these data, dislike of
school has been included in models of entry into parenthood specifically and repre-
sents the first such research to do so to the author’s knowledge.
Socioeconomic Factors
Socioeconomic factors have been implicated in a wide range of studies of the pre-
dictors of early parenthood and are pervasive in most research, either explicitly or as
background controls (Bynner et al., 2000, Ermisch and Pevalin, 2003b, Ekert-Jaffe
et al., 2002). Usually socioeconomic factors include income, social class and tenure.
In this research, while all three spheres are tested to a certain extent, tenure is found
to be the strongest predictor.
All three factors are meant to capture an effect of disadvantage and to predict
the probability of poor labour market success. Schoon and colleagues’ research
provides a useful framework with socioeconomic background being “one of the
main predictors of cognitive development, which provides the underpinnings of aca-
demic achievement upon which much success in later life depends” (Schoon et al.,
2002, p. 1486). However, the fact that socioeconomic factors retain significance
and sometimes outweigh the effect of educational predictors of early parenthood
suggests that socioeconomic factors are capturing an element of labour market dis-
advantage not completely accounted for by educational factors. Other pathways
through which socioeconomic factors can operate include lowered personal, social
and sexual negotiation skills, limited access to healthcare, lack of positive role mod-
els and living in dangerous environments (Singh et al., 2001). In this research,
socioeconomic factors are found to be instrumental in distinguishing some defi-
nitions of ‘early’ parenthood and are also found to have differential effects between
motherhood and fatherhood.
Demographic Factors
An unstable home life is consistently identified as a predictor for the timing of moth-
erhood. This can include a history of being in social services care, sexual abuse,
parental divorce and parental mental health problems (Kiernan and Hobcraft, 1997).
In this research, against other background controls, an indicator of unstable

parental structure (covering divorce, separation, death and foster care) is found to
be a relatively weak predictor of early parenthood. A far more consistent predictor
is the age of cohort members’ parents at first birth. This is consistent with other
literature which finds a cyclical pattern of the timing of motherhood. The Social
Exclusion Unit reports that sisters and daughters of teen mothers are six times more
likely to become pregnant at an early age (Social Exclusion Unit, 1999). In their
analysis of BCS70 data, Ermisch and Pevalin find that women born to teenage or
young adult mothers (20–23 years old) are around two and a half times more likely
to have a teenage birth themselves (Ermisch and Pevalin, 2003a). This evidence is
also consistent with studies of early fatherhood (Berrington et al., 2005).
Ethnicity is also deemed an important predictor of the timing of motherhood,
with non-white women usually at risk of experiencing early births in the UK and
US (Social Exclusion Unit, 1999, Singh et al., 2001; Robson and Berthoud, 2006).
Unfortunately in these data, it is not possible to investigate ethnicity fully because of
the low numbers of cohort members from ethnic minorities included in the NCDS
and BCS70.5
Behavioural and Philoprogenitive Factors
In a large portion of the literature on early parenthood, and on teenage motherhood

in particular, entry into parenthood is viewed as the outcome of a risk-taking per-
sonality, with early parenthood being linked to other risk-taking behaviours such
as smoking, drug taking and alcohol consumption (Birch, 1992; Social Exclusion
Unit, 1999; Jaffee et al., 2001). This would appear to support a theory that early
parenthood may be equated with unplanned parenthood – unprotected sex being a
risk factor and early parenthood the outcome. While for the more educated cohort
members and for the BCS70 cohort as a whole compared to NCDS, access to abor-
tion as recourse would be easier, this would not have been an option for many cohort
members. The assumption of unplanned parenthood could be viewed as a criticism
of the theory of the importance of opportunity costs as a motivation of early parent-
hood. However, the choices once pregnant may instead be viewed as a reflection of
this. In this research, proxy measures for family-building orientation are examined
as predictors of early parenthood.
In addition, this research examines some of the drivers that could influence
family-building intentions through analysing indices of behaviour. Behavioural
characteristics may operate directly on a pathway to early parenthood through influ-
encing family-building intentions or may operate through lack of self esteem or poor
sexual negotiation skills (Jaffee et al., 2001; Hobcraft and Kiernan, 2001; Kendall
et al., 2005; Birch, 1996). In particular in their study of early fatherhood, Jaffee and
colleagues found a history of conduct disorder to be a significant predictor of early
5 The later Millennium Cohort Study over-sampled within areas with a high ethnic minority popu-
lation so that ethnic group effects could be researched. See chapter in this volume by Hawkes et al.
86 D. Kneale
fatherhood (Jaffee et al., 2001) which may be linked to more general social dysfunc-
tion and withdrawal. Hobcraft and Kiernan (2001) examined anxiety, aggression and
restlessness as drivers of early motherhood. Similar indicators using Rutter score
measures are used later in this research, where it is found that behavioural and var-
ious other measures do not account for family-building orientation, suggesting this
orientation may be linked to other measures such as contextual predictors. Further
work on contextual predictors investigates this hypothesis further (Kneale, 2008).
While it is not within the scope of this chapter to examine all these factors simul-
taneously (data constraints being one reason), the results presented in this chapter do
successfully unpack some of these processes occurring in childhood. In particular,
given the rich data source used in these analyses, a number of these predictors can
be compared across definitions of ‘early’ parenthood, across gender and, to a certain
extent, across time. These results are discussed in detail in later sections once the
aims of the research have been clarified, the data sources have been discussed, and
the research methodology has been introduced.
Research Questions and Data

The aim of the rest of this chapter is to examine, in brief, the following research
questions:
• Do teenage parents differ in their backgrounds to parents in their early twenties?
• Do the predictors of early motherhood vary from those of early fatherhood?
• How have the predictors of early parenthood changed between two British
cohorts?
As these questions are wide ranging, full results from every model are not pre-
sented comprehensively. This chapter highlights some of the main findings from
the results of the modelling exercise and to signpost the reader to potential con-
siderations when researching patterns of transition to early parenthood. In addition,
this chapter only presents results from parsimonious models. While a full range
of predictors, listed later, were tested; only those significant in models containing
other controlling factors are discussed. We return to the methodology once the data
sources have been considered.
This research uses two of the four British birth cohort studies – prospective lon-
gitudinal studies following the lives of individuals born during 1958 (the National
Child Development Study (NCDS)) and 1970 (the British Cohort Survey (BCS70)).
Recent papers have outlined the history of these studies and some of their most
prominent findings (Plewis et al., 2004; Bynner and Joshi, 2007; Ferri, 1993; Elliott
and Shepherd, 2006; Power and Elliott, 2006). Both studies prospectively followed
individuals and were essentially a census of all born in one week in 1958 and 1970.
For the NCDS, further data collection has occurred at ages 7, 11, 16, 23, 33, 41/42
and 46 years. Data from BCS70 cohort members were collected at ages 5, 10, 16,
26, 29/30 and 34 years. Data relevant to a wide spectrum of disciplines have been
collected through these studies and combined, over 1,200 publications have doc-
umented some of the major findings of these studies (Elliott and Shepherd, 2006;
Power and Elliott, 2006).
This research exploits the breadth of these data in examining a number of child-
hood factors that are thought to predict early parenthood in the literature and
explores some new predictors. NCDS was a pioneer in asking men to report their
birth histories. It is nevertheless acknowledged that fatherhood is likely to be under-
reported (Rendall et al., 1999). This is particularly likely if the father is not living
with the mother of his child and again particularly likely for the youngest moth-
ers and fathers (Greene and Biddlecom, 2000; Rendall et al., 1999). In this chapter,
live births reported by cohort members are examined, excluding stillbirths and other
fertility outcomes. Those still pregnant or those who have fathered a pregnancy not
carried to full term and, inevitably, births which were not reported by either men or
women are excluded.
Fertility histories were collected from NCDS cohort members from age 16
onwards; while full histories for BCS70 members were not collected until 30 years.
For the NCDS cohort full fertility histories were collected at 23, 33, 42 and 46
years.6 In total, over 14,000 fertility histories are available for NCDS while for
BCS70, almost 12,000 fertility histories are available. Analysis of transition to first
parenthood using survival curves suggest that both cohorts are representative in
terms of fertility patterns when compared to Office for National Statistics estimates
(Kneale, 2008; Office for National Statistics, 2007).
Methodology
Even in the 12 years between cohorts, age at first parenthood has increased sig-
nificantly while the teenage parent population decreased. Among women, 13%
of NCDS cohort members became teenage mothers dropping to 10% among the
BCS70; while among men, 4% of the NCDS cohort were teenage fathers, drop-
ping to a mere 3% among BCS70 members. Event history treatment of the data
put the point at which the first 25% (lower quartile) of the whole NCDS cohort
had entered into motherhood at 22 years 2 months. For BCS70, this point had risen
by over 1 year 6 months to 23 years 11 months. For fatherhood, the pattern was
even more startling, with the age limit of the first quarter to enter fatherhood rising
from 24 years 11 months among NCDS to 27 years 1 month among BCS70. Such
a rise would give the first indications that a definition of ‘early’ that is grounded
against normative patterns of parenthood would respond to changes over time. This
is reflected in Fig. 5.1, which also demonstrates the unequal shift in the rise in age
at first birth mentioned above.
6 Some observations have been artificially truncated at age 23 years, while information for a small
minority of others has not been used in this analysis.
88 D. Kneale
100
Percentage entering parenthood

75
50
25
NCDS male NCDS female BCS70 male BCS70 female

0
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Age
Fig. 5.1 Entry into parenthood up to 30 years by gender and cohort: NCDS and BCS70 cohorts
(Source: Author’s analysis)
The assumption often made in the literature that teenage parents significantly
differ from those in their twenties, was tested through piecewise linear regression.
Using predictor and outcome information in univariate analysis, this technique only
located break points (the points at which the trend differs significantly) well into
the twenties, and even into the thirties in the case of BCS70 fatherhood (Kneale,
2008). This cast doubt on the usefulness of threshold of early parenthood at age
20. However, using break points derived from piecewise regression may be a poor
choice if there is a need to make comparisons across genders and cohorts.
Another approach is to explore a relative definition of ‘early’. Two of these are
tested here: firstly, ‘very early’ representing the first 12.5% of the cohort (by gender)
to enter into parenthood, and secondly, ‘early’ as the first 25% to enter parenthood;
besides the absolute dividing line at age 20.7 These are displayed in Table 5.1.
Binary logistic regression models are used to examine the strength of predic-
tors using these definitions of early and employing a backward elimination method.
These estimate the probability of becoming an early parent versus not becom-
ing an early parent after accounting for a number of known predictors. Those not
becoming an early parent include those who became a parent at a later stage, who
remained childless, or who left the studies and were lost to follow up.8 Binary logis-
tic as opposed to multinomial logistic regression models are used because of the
uncertainty between the categories that represent ‘not becoming an early parent’.
Event history models are used to examine the effect of predictors of the timing
7 Information for models of very early motherhood in NCDS are not presented because this dis-
tinction corresponds very closely with the teenage definition.
8 Censoring began at 23 years onwards for some NCDS cohort members and 30 years onwards for
BCS70.
Table 5.1 Age at entry into parenthood at each quartile and levels of childlessness at last obser-
vation: BCS70 and NCDS cohorts
First First
12.5% to First 25% 75% to
become to become First 50% become Childless
parents parents to become parents at last
(very (lower parents (upper observation
Gender Cohort early) quartile) (median) quartile) (%)
♂ NCDS 22 years 24 years 29 years 38 years 20.8

5 months 11 months 5 months 2 months
BCS70 23 years 27 years 33 years – 41.6
10 months 1 month 2 months
♀ NCDS 19 years 22 years 26 years 32 years 15.6
10 months 2 months 6 months 5 months
BCS70 20 years 23 years 29 years – 27.4
9 months 11 months 2 months
Source: Author’s analysis.
of parenthood over the ‘early’ years, but these results are not presented in detail
here. The focus of results presented here is the effect of childhood factors (mea-
sured up to the age of 16 years) as predictors of early parenthood. While it is
recognised that early adult predictors may increase and actually overtake child-
hood factors in their potency, given their proximity to the event under study, they
could introduce problems of reverse causality and are left beyond the scope of
this study.
Summary of Results
As can be observed in Table 5.2 for the NCDS cohort and Table 5.3 for the BCS70
cohort, representing only a summary of significant factors in full models, a complex
and diverse formula of states and factors help to predict entry into early parenthood.9
A number of significant trends stand out that not only illuminate the predictors of
early, as opposed to teenage parenthood, but also challenge conventional wisdom.
In particular, of note are: housing tenure as a predictor over social class for both
cohorts; the consistency of disliking school as a strong predictor for both sexes; and
the case of age 20 as being a break in the continuum of early fatherhood character-
istics; as being distinct in these data. In these tables, the area under the ROC curve
9 The results in these models represent the most parsimonious fitting model for predicting each
definition. Other predictors tested but not found to be significant against other controls were as
listed: NCDS: Parental Structure at ages 0, 7 and 11; other education test scores at ages 7 and 11;
Cohort Participation. BCS70: Parental Structure at ages 0 and 10; other education test scores at
ages 5 and 10 (and 16); receipt of Unemployment and Sickness benefits and School Attendance at
age 11 years.
90 D. Kneale
Table 5.2 Summary of significant predictors of early parenthood: NCDS Cohort
♂ ♀
Very early
NCDS Early (First 25%) (First 12.5%) Teenage Early (First 25%) Teenage
Father’s social
class
Tenure
Personality
(Age 16)
Intentions (Age
16)
Maths score
(Age 16)
Reading score
(Age 16)
School dislike
(Age 16)
Attendance
(Age 16)
Parental age at
first birth
Parental
structure
age 16
Sample size 2,367 3,278 2,859 1,955 2,784
Area under 0.729 0.744 0.793 0.794 0.822
ROC curve
(Receiver Operating Characteristic) is calculated. This is reflective of the accuracy

of the model – for example 0.75 is reflective of the fact that we would correctly esti-
mate an observed early parent to have a higher probability of early parenthood than
someone who wasn’t an observed early parent 75% of the time. Generally, a value
above 70% is considered acceptable and a value above 80% excellent (Hosmer and
Lemeshow, 2000).10
Direct measures of childhood socioeconomic conditions here signify the social
class environment, tenure environment and receipt of unemployment or sickness
benefit at age 10 (BCS70) and age 11 (NCDS). Social class environment measures
the number of times a cohort member’s father has been identified as belonging to
non manual classes (NCDS) and Social Class I and II (BCS70). This is then cate-
gorised into all observations, some observations or never recorded. In order to max-
imise sample size, while incorporating some measure of social class mobility within
10 Allmodels also passed the Hosmer-Lemeshow goodness of fit test which is a comparison of
observed and expected frequencies of events across strata and is also a good indicator of redundant
variables in models. In addition, likelihood ratio tests were also used to assess the significance of
individual variables.
Table 5.3 Summary of significant predictors of early parenthood: BCS70 Cohort
♂ ♀
Early Early
(First Very early (First Very early
BCS70 12.5%) (First 12.5%) Teenage 12.5%) (First 12.5%) Teenage
Father’s social
class
Tenure
Personality
(Age 5)
Personality
(Age 10)
Personality
(Age 16)
Intentions
(Age 16)
Parental education
English
Vocabulary
Score (Age 5)
Maths score
(Age 10)
Reading score 1 1
(Age 10)
Matrices score
(Age 10)
School dislike 1 1
(Age 16)
Parental age at first
birth
Parental structure
(age 5)
Parental structure
(age 16)
Unemployment/
sickness
benefits
(Age 16)
Truancy (Age 16)
Sample size 3,206 2,220 3,712 2,704 2,809 3,232
Area under ROC 0.709 0.763 0.792 0.781 0.797 0.807
curve
1
Reading Age 10∗ School Dislike Interaction Term
the parental generation, this variable did not differentiate by the number of obser-
vations recorded for a cohort member. For example, a cohort member in a certain
social class at age 10 years but not observed another time would be classified as
having all observations in that social class. This decision was taken after it was
92 D. Kneale
observed that transition between waves remained relatively stable after the birth
and first waves of data collection, which in themselves have high wave response
rates. In an acknowledgement that this variable is highly dependent of participation
at childhood waves of data collection, a control variable was tested that measured
participation, although was found to be insignificant for the most part.
A similar strategy was adopted for tenure (with a breakdown given later) while
the benefits variable was a binary variable reflecting receipt of state unemploy-
ment and sickness benefits at a point in childhood. Measures of behaviour represent
components derived from principal components analysis of Rutter Score measures
(Rutter, 1967) that are common to both cohorts. These are an established set of mea-
sures to monitor signs of behaviour disorders in children and teenagers including
aggression, disobedience and nervousness. These differ in their importance between
models, with an index of aggression/misbehaviour particularly important for defini-
tions of BCS70 fatherhood).
Teenage Fatherhood as a Choice Versus Teenage Motherhood

as an Adaptation: Breaks in the Continuum
As mentioned previously, the focus on teenage parents has seemed disproportionate
given their low, and diminishing, prevalence and a large part of the investigation
has been to examine if, in terms of predictors, their characteristics vary significantly
from people who became parents in their early 20s. This was not generally found
to be the case. However, a peculiarity that does stand out in terms of the results
is the weak roles that direct measures of socioeconomic status have in the case
of teenage fatherhood. This did not apply to more extensive definitions of early
fatherhood, and appears to break the continuum between teenage and other forms of
early fatherhood. This phenomenon is present in both cohorts, with a total absence
of direct socioeconomic measures registering in the models of teenage fatherhood.
In fact for the NCDS model of early fatherhood, few conventional predictors are
found to predict entry into this state.
All variables included in the fatherhood models presented in Tables 5.2 and 5.3
have significant predictive power. Model fit statistics suggest that the models for
predicting early fatherhood provide a weaker framework of predictors than teenage
fatherhood. While it would be expected that early fatherhood would constitute a
more diverse group of fathers than teenage fatherhood; these results suggest that
those joining the early fatherhood group (that already includes teenage fathers) are
actually a group that is governed by a significantly different set of predictors than
those of teenage fathers, to the point where modelling two distinct populations as
a joint category produces a relatively poorer fitting model.11 This is demonstrated
11 Using the Hosmer-Lemeshow test; although these results appear to be dependent on the way that
predictions are grouped (Harrell, 2001) and are not presented here in favour of ROC curve results,
that are less responsive to this.
Table 5.4 Goodness of fit statistics (area under ROC curve): binary definitions of early
fatherhood
Model excluding Model excluding

direct childhood behavioural and
Definition of socioeconomic philoprogeneity
Cohort fatherhood measures measures∗ Full model N
NCDS Early 0.699 0.708 0.722 2,256

Very early 0.702 0.719 0.728 3,205
Teenage n/a 0.728 0.758 3,263
BCS70 Early 0.697 0.708 0.709 3,206
Very early 0.759 0.750 0.763 2,220
Teenage n/a 0.735 0.792 3,712

∗
Includes Dislike of School
in Table 5.4, which shows the impact of removing different sets of predictors, with
higher values representing a better prediction. Socioeconomic factors have greater
predictive influence in more inclusive early fatherhood models but are found to be
insignificant in teenage models.
The results suggest that while transition to fatherhood in the early 20s is an adap-
tation to socioeconomic factors and probably more proximal measures of socioeco-
nomic status; fatherhood under 20 is associated more with behavioural, motivational
and possibly role model elements. In the case of fatherhood therefore, the contin-
uum appears to move from different sets of predictive factors (from behavioural to
socioeconomic) that govern transition, while for motherhood it appears more of a
continuum moving gradually in terms of strength of prediction of the same variable
groupings.
In particular, the model for teenage fatherhood may be more of a model reflecting
sexual behaviour among teenagers than conscious transition to fatherhood. Unfor-
tunately, it is difficult to assess this supposition given that there are few studies
that collect behavioural measurements with information on sexual behaviour, with
the latter not collected in the NCDS or BCS70. The inclusion of mother’s age of
finishing continuous education in the BCS70 teenage fatherhood model and the sig-
nificance of mother’s and father’s age at first birth in the BCS70 and NCDS models
respectively may represent a socioeconomic dimension or may well reflect parental
input (including an element of having a role model), which may be moderators of
teenage sexual behaviour.12
The BCS70 teenage fatherhood model finds that having a high score for an
aggressive, disobedient and destructive nature at age 16 years to be significant.
In fact, those with a score in the highest quartile for this component were 4.6 times
(CI: 1.9–11.2) more likely to become teenage fathers than those with a score in the
lowest quartile. For BCS70, dislike of school at age 16 (discussed later) was found
12 Parental environment is examined in depth in other parts of the wider research Kneale (2008).
94 D. Kneale
to be highly significant. Those who disagreed in full with a statement on disliking

school were 83% less likely to become teenage fathers in BCS70 (odds ratio (OR):
0.17; CI: 0.07–0.43). In NCDS, those who were uncertain as to the ideal age at
which they wanted to start a family were 85% (OR: 0.15; CI: 0.03–0.44) less likely
to become teenage fathers than those who thought it ideal to start a family at age
16–19 years. While some of these factors were significant in more inclusive defi-
nitions of ‘early’ fatherhood; it was the fact that these were more prominent than
socioeconomic measures that defines teenage fatherhood from early fatherhood in
these models.
Teenage motherhood models show a greater continuum between teenage and
other definitions of early motherhood. In addition, as might be expected, a greater
range of covariates have been identified as significant for early motherhood than
early fatherhood, which may indicate a less random social profile of young moth-
ers. The stronger socioeconomic component suggests that early motherhood (or its
avoidance) is more of an adaptation to economic circumstance than is the case for
early fatherhood, which appears to be governed by behavioural rather than economic
factors. This could be because fathering a child as a teenager is less likely to involve
co-residence with the child than is the case for teenage motherhood and is likely to
have less of a direct economic consequence. This is a temporal state however and
has been observed to change over time where some teenage fathers take up resi-
dence with their children a few years after birth (Clarke et al., 2000). In essence as
has been stated, in the case of teenage fatherhood, often but not always it may be a
case of modelling sexual behaviour and partnership patterns rather than conscious
behaviour.
For both cohorts, all three definitions of early motherhood are governed by
similar sets of predictors. Universal predictors of young motherhood across both
cohorts include tenure, dislike of school at age 16, and cohort members’ mother’s
age at first birth. As a generalisation, the effect of covariates wanes slightly
between teenage and early motherhood models, with early motherhood models
representing a more diluted category, as a greater number of significant predic-
tors are included in the more inclusive early motherhood models. However, in
summary, it would appear that a continuum does exist between these definitions
of early motherhood, given that the predictors appear to maintain their effect
(Fig. 5.2).
The Potent Effect of Disliking School on Entry

into Early Parenthood
Dislike of school itself has been examined as a predictor of early pregnancy with
mixed results (Bonell et al., 2005). This is, to the author’s knowledge, the first
attempt to include ‘school dislike’ in models of entry into parenthood. Dislike
of school is found to be a potent predictor of early entry into parenthood, not
only of teenage motherhood, but also across most definitions of early parenthood
4
Teenage motherhood NCDS
Teenage motherhood BCS70
Early motherhood NCDS
Early motherhood BCS70
3
Odds of early motherhood
0
occupation
Only council
housing
Some council
No owner occupation
Other
Each additional
Doesn’t dislike
school
year
Dislikes school
somewhat
Mixed owner
Tenure (baseline owner occupation) Age of School dislike

CM’s (baseline strong
mother at dislike of school)
first birth
Fig. 5.2 The odds of becoming an early or teenage mother: the effects of selected covariates from
models with full controls (Source: Author’s analysis)
See Tables 5.1 and 5.2 for a full list of controls and sample sizes
The NCDS scale has been mapped into a three point scale for the purposes of Fig. 5.2 and Table
5.4, but maintains a 5 point scale elsewhere in the models
(including fatherhood). Furthermore, given that dislike of school is measured in

a fairly consistent way across both cohorts, its effect appears to be increasing
over time.
In these data, aversion to school has been measured as the agreement or dis-
agreement with the statement ‘I do not like school’. This was measured at age 16 in
both cohorts – in the NCDS as a five point scale while in the BCS70 this was as a
three point scale. As the BCS70 cohort had a particularly low response of the age
16 sweep (Plewis et al., 2004), a missing category was created, which was repli-
cated for NCDS for comparability purposes. Among NCDS boys, 17% reported a
strong dislike of school with 18% in BCS70, while the levels for girls was slightly
lower at 15 and 16% respectively for NCDS and BCS70. While the level of dislike
96 D. Kneale
Table 5.5 The effect of school dislike on early parenthood: odds ratio results from parsimonious
main effects models NCDS and BCS70
School dislike (Baseline: strong dislike of school)

Dislikes school Does not dislike Item non-
Cohort Definition somewhat school response∗∗∗
NCDS Early fatherhood 0.842 0.706∗ n/a

Very early fatherhood 0.816 0.522∗∗ n/a
Teenage fatherhood Not significant in full model
Early motherhood 0.744 0.710∗ n/a
Teenage motherhood 0.517∗∗ 0.411∗∗ n/a
BCS70 Early fatherhood 0.658∗ 0.592∗∗ 0.994
Very early fatherhood 0.520∗ 0.329∗∗ 0.770
Teenage fatherhood 0.209∗∗ 0.170∗∗ 0.520∗
Early motherhood 0.593∗∗ 0.507∗∗ 0.951
Very early motherhood 0.557∗∗ 0.503∗∗ 0.561
Teenage motherhood 0.502∗∗ 0.394∗∗ 0.656
∗∗
p < 0.01; ∗ p < 0.05 ∗∗∗ As the item non-response for dislike of school remained fairly low for
NCDS (accounting for less 20%); this category was not created.
of school has remained fairly constant, the effect may be growing. The results for
school dislike in models containing a full set of significant predictors are presented
in Table 5.5.
Using those who dislike school as a baseline, it appears that even partial dis-
agreement with the statement ‘I do not like school’ is highly protective against
early parenthood, particularly for the BCS70 cohort. In the case of teenage father-
hood, even partial disagreement leads to a 79% reduction in the odds of becoming a
teenage father. In lognormal event history models, for entry into fatherhood between
the ages of 16–23 years among BCS70 males, partial disagreement with the state-
ment leads to a 14% increase in the time spent childless (time ratio (TR): 1.14; CI:
1.08–1.21) with full disagreement leading to an 18% increase (TR: 1.17; CI:
1.09–1.25). Dislike of school produces the largest coefficients in this model, and
maintains this effect when the observation time is extended to 30 years, overshad-
owing socioeconomic effects.
Dislike of school in the case of BCS70 teenage fatherhood outweighs the sig-
nificance of socioeconomic measures, while it is significant alongside socioeco-
nomic measures in other models. Such a finding has not been replicated in the
other few studies of dislike of school (Bonell et al., 2005). In addition, in these
data dislike of school is found to be significant alongside measures of educational
achievement, and again, overshadowing their impact in several cases. The effects
were amplified with educational attainment in BCS70, so that high achievement
and liking school were mutually reinforcing protective effects against early mother-
hood. From a policy perspective, this finding is highly significant and suggests that
where early parenthood is viewed as problematic, that school based interventions to
improve engagement could have measurable results. Also of interest in these data
and possibly related to dislike of school is the prominence of family building inten-
tions and values as predictors of early parenthood, explored in the next subsection.
Early Parenthood as Planned Parenthood
Philoprogenitive tendencies (tendencies orientated towards children) might appear

as obvious predictors of early parenthood, but they have actually received little atten-
tion. This is, possibly because of the assumption that very early parenthood is the
result of unplanned pregnancy (Social Exclusion Unit, 1999), because these tenden-
cies are subject to revision, because of reliance on socioeconomic theories of fer-
tility or because of the lack of prospective data. In fact, family building intentions
have been more the focus as concerns of postponement and childlessness (Simp-
son, 2006; Kneale and Joshi, 2008; Berrington, 2004) than early fertility, although
some focus has been made in studies of teenage motherhood. One example is the
index of ‘positive orientation towards early motherhood’ coined by Afable-Munsuz
and colleagues, composed of a series of statements about the value of children
where a latent desire for children was found among those becoming early mothers
(Afable-Munsuz et al., 2005). However, a drawback of Afable-Munsuz’s study is
that it that its sample was a narrow population of African-American young women
who were existing patrons of family planning services. East and colleagues also
found similar results within a narrow high risk population (East et al., 2006). In the
present data, information has been collected in a less selective way, for two cohorts
of both genders.
NCDS cohort members aged 16 year olds were asked about the age at which
they would ideally start a family, with responses grouped by age and also a category
formed for those who intended to be childless. The modal response category for both
sexes was 22–25 years and actually only a small number of cohort members chose
the youngest category of 16–19 years13 (2% of boys and 3% of girls; those choosing
under 22 years had higher numbers; 13% of boys and 18% of girls). Almost 10%
of females and 3% of males choosing the childless category became teenage par-
ents. BCS70 cohort members were asked about the importance of children in their
own life to come at the age of 16 years on a three point scale. Females were most
likely to answer that children mattered very much while males had higher levels that
answered that children only mattered somewhat. While both measures are proxies of
intentions, both will give an indication of family building preferences as predictors,
and are referred to as family building intentions from this point forward.
These data find that philoprogenitive tendencies are generally a better predictor
of early motherhood than early fatherhood. Only the models for NCDS early and
teenage fatherhood find these to be a mildly significant predictor. In the case of
13 This is a combined category in the data of 16–17 and 18–19. There was no option to choose any
earlier.
98 D. Kneale
Table 5.6 The effect of intentions, as measured by importance of children, on early parenthood:
odds ratios from parsimonious main effects models
Early motherhood Teenage motherhood
Baseline (children very important)

Children somewhat important 0.540∗∗ 0.660
Children not important 0.524∗∗ 0.402∗∗
Item non-response 0.672∗ 0.884
∗∗
p < 0.01; ∗ p < 0.05
models of early motherhood among NCDS cohort members, the age at which the
cohort member regards as the ideal age at first birth operates in a non-linear fash-
ion (and is modelled as a categorical variable), with those uncertain about the ideal
age or those selecting 26–30 years having the lowest odds of becoming early moth-
ers. Among BCS70 models of early motherhood, the importance of children is a
significant predictor of teenage and early forms of motherhood with those deeming
children to be unimportant half as likely to become early or teenage mothers than
those deeming children to be very important. The results are presented in Table 5.6
and Fig. 5.3.
These data confirm that becoming an early mother is a process based upon more
than just economic adaptation. For many young mothers, becoming an early mother
is grounded in a desire to enter motherhood at an early age and a high value placed
upon children. These factors remain significant even after controlling for socioe-
conomic factors, educational achievement, behavioural factors and more relevant
perhaps, school dislike. Early motherhood is often taken for granted solely as a
rational adaptation to economic factors, with poor labour market prospects leading
to some reduced opportunity costs in having an early child (Ermisch and Pevalin,
2003b; Hoem, 2000). These data would demonstrate that even after accounting for
Teenage motherhood NCDS Early motherhood NCDS

2
Odds of early motherhood
Fig. 5.3 The effect of 1

intentions, as measured by
ideal age at first birth, on
early parenthood: odds ratios
from parsimonious main
effects models 0
16–19
20–21
22–25
26–30
(Source: Author’s analysis)

Uncertain
Children
Item non-
Over 30
response
No to
Hatched lines represent

confidence intervals for
teenage motherhood model
factors influencing labour market opportunity costs (including educational factors

and dislike of school), that a social norm influencing childbearing orientations
remains for women. Should these preferences and social norms be consolidated
on a community or neighbourhood level, it would be better to think of a teenage
or any early mother as being “an emerging adult participant in active multigener-
ational social networks, than as a rebellious adolescent set apart from her elders”
(Geronimus, 1997, p. 418).
The Dominant Role of Housing Tenure Over Social Class
A striking find in this analysis is the dominant role that housing tenure has in pre-
dicting early parenthood, alongside or even eclipsing the role of social class. Results
from models of fatherhood are presented in Table 5.7 while Fig. 5.2 shows some
results for motherhood. Housing tenure is one of the few predictors that moderates
entry into nearly all forms of parenthood in both cohorts (with the exception of
teenage fatherhood) in both event history and binary logistic modelling strategies.
One could take the view that housing tenure represents a proxy for the type of neigh-
bourhood. It could also be a direct measure of childhood socioeconomic conditions.
The significance of tenure contributes to an emerging theme of the importance of
contextual factors (Kneale, 2008).
Table 5.7 shows that tenure has an equal and sometimes greater association
with entry into early fatherhood than it has with entry into very early fatherhood.
This is mirrored to some extent in BCS70 motherhood, but is contrary to most
other results in this research that have found that common predictors among def-
initions have waned with more inclusive definitions of young parenthood. What
this analysis, backed up by results from Event History models, adds to the story
of early transition to parenthood is that housing tenure in childhood has effects
Table 5.7 The effect of tenure on early parenthood: odds ratio results from parsimonious main
effects models NCDS and BCS70 of fatherhood
Tenure (Baseline: owner occupation)

Mixed owner Some council, no
Definition of occupation Only council owner occupation
Cohort Fatherhood tenure tenure tenure Other
NCDS Early 1.52∗∗ 1.83∗∗ 1.52 1.67∗

Very early 1.34 2.07∗∗ 2.33∗∗ 1.55
Teenage Not significant in full model
BCS70 Early 1.14 1.49∗∗ 1.29 1.31
Very early 1.23 1.50∗ 0.78 1.08
Teenage Not significant in full model
∗∗
p < 0.01; ∗ p < 0.05
100 D. Kneale
lasting beyond conventionally defined adolescence and well into the twenties. In
addition, this analysis shows that in certain cases, experiencing social housing, even
if accompanied by upward movement into owner occupation, which may have been
facilitated by the right-to-buy scheme for BCS70, has significant and lasting effects.
This may well be as related to retaining community ties such as maintaining school
or peer groups, or the lasting legacy of childhood poverty. These conclusions mir-
ror those of other research that has examined childhood tenure and adult outcomes
(Feinstein et al., 2008).
Conclusions: Building a Picture of Early Parenthood

Patterns in the UK
In this chapter, a definition of early parenthood bounded by the age at which one
quarter of a cohort had become parents was explored. For women born in 1958,
this was just over 22; for those born in 1970 it had risen to just over 24. The cor-
responding figures for men were around 25 and 27. On this definition most ‘early’
childbearing is no longer confined to teenagers who account for less than 13 and
10% of women and 4 and 3% of men among the NCDS and BCS70 cohorts.
Early parenthood has been viewed through a wide prism in this research and as
a result, this chapter illuminates several themes in the story of young parenthood in
the UK. One of the first themes identified was the inadequacy of drawing the line
at age 20. This inadequacy was revealed through univariate techniques, reinforced
by multivariate exploration of predictors. However, the story was not straightfor-
ward. While teenage motherhood and early motherhood were described by similar
processes and were essentially viewed as being on a continuum, the even smaller
group of males reporting fatherhood as teenagers appeared to be a distinct group. In
particular, the lack of association between teenage fatherhood and socioeconomic
factors was hypothesised as reflecting teenage male sexuality and risky behaviour
rather than planned fatherhood. There are greater means and incentives for women
to avert the consequences of an unplanned pregnancy than males. In the case of
early fatherhood, there was a discontinuity as the earliest (teenage) fatherhood was
associated with behavioural factors and movement towards socioeconomic factors
with age.
Looking at particular predictors of early parenthood revealed some novel results.
The dislike of school was identified as being pervasive in most models of early
parenthood for both cohorts and genders. Having been previously linked to early
pregnancy alone, and then only in models excluding socioeconomic predictors,
its continuing strength in predicting early parenthood was surprising. The path-
way behind this finding is unknown. This could reflect reduced opportunity costs
of entering parenthood early, as is hypothesised in literature on early parenthood
(Ermisch and Pevalin, 2003b); or alternatively be more of an indicator of con-
textual factors such as peer group, school or community effects; or a combina-
tion of both. The significance of philoprogeneity, as measured by ideal age at first
birth and importance of children in the NCDS and BCS70 respectively, even when
controlling for a battery of other predictors, may again signify the importance of
contextual factors. Finally, the strong predictive power of housing tenure and the
observed effect of experience of poverty threw into question traditional measures of
social class as sole measures of socioeconomic circumstance.
Cross-cohort comparisons throughout have revealed slight nuances between the
cohorts in the way these predictors operate. However, the overwhelming theme is
one where the issue of comparability between cohorts can only really be made when
relative measures are used. The inadequacy of using a teenage definition of early
parenthood becomes particularly acute for BCS70 where the ‘absolute’ definition of
early applies to an increasingly marginalised group. Of note as well is the increased
potency of some predictors among the BCS70 cohort across all models suggesting
that as a whole, social polarisation in age at first birth is forming a stronger discourse
in the more recent cohort.
The results presented here have composed a picture of transition to early parent-
hood as being governed by numerous processes. In particular they have identified
numerous pathways to early transition that vary from conventional predictors and
definitions of early parenthood. These investigations have also revealed avenues for
future work in investigating contextual predictors that are under investigation in fur-
ther research (Kneale, 2008). While a number of significant predictors have been
found to govern early parenthood, model fit statistics suggest that many more are
yet to be found under the assumption that the timing of parenthood is more than just
a random event.
Acknowledgments I would like to acknowledge the contribution of Heather Joshi and Jane Elliott
throughout this project both in editorial support and with analytical issues. I am very grateful
for this support. I would also like to thank Brian Dodgeon for his help in solving data issues
encountered in this chapter. Finally, I would like to acknowledge UPTAP and the ESRC for their
financial support as well as all those involved with the NCDS and BCS70 studies including the
cohort members themselves.
References
Afable-Munsuz, A., Speizer, I., Magnus, J. H. and Kendall, C. (2005) A positive orientation toward
early motherhood is associated with unintended pregnancy among New Orleans youth, Mater-
nal and Child Health Journal, 10: 265–276.
Arai, L. (2003) Low expectations, sexual attitudes and knowledge: explaining teenage pregnancy
and fertility in English communities. Insights from qualitative research, Sociological Review,
51: 199–217.
Berrington, A.M. (2004) Perpetual Postponers? Women’s, men’s and couple’s fertility intentions
and subsequent fertility behaviour, Population Trends, 117: 9–19.
Berrington, A.M., Cobos Hernandez, M.I., Ingham, R. and Stevenson, J. (2005) Antecedents and
outcomes of young fatherhood: longitudinal evidence from the 1970 British Birth Cohort Study,
S3RI Applications and Policy Working Papers, Southampton Statistical Sciences Research
Institute, Southampton University, Southampton.
Birch, D. (1992) Are You My Sister Mummy, Youth Support, London.
Birch, D. (1996) The Child that Rocks the Cradle, Youth Support, London.
102 D. Kneale
Bonell, C., Allen, E., Strgange, V., Copas, A., Oakley, A., Johnson, A. and Stephenson, J. (2005)
The effect of dislike of school on risk of teenage pregnancy: testing of hypotheses using
longitudinal data from a randomised trial of sex education, Journal of Epidemiology and Com-
munity Health, 59: 223–230.
Bonell, C., Fletcher, A. and McCambridge, J. (2007) Improving school ethos may reduce substance
misuse and teenage pregnancy, British medical Journal, 334: 614–616.
Bynner, R.J. and Joshi, H. (2007) Building the evidence base from longitudinal data: the aims,
content and achievements of the British Birth Cohort Studies, Working Paper, Centre for Lon-
gitudinal Studies, Institute of Education, University of London, London.
Bynner, J., Joshi, H. and Tsatsas, M. (2000) Obstacles and Opportunities on the Route to Adult-
hood: Evidence from Rural and Urban Britain, The Smith Institute, London.
Clarke, L., Joshi, H. and Di Salvo, P. (2000) Reports and records of mothers, fathers and children
compared, Population Trends, 102: 24–33.
East, P.L., Khoo, S.T. and Reyes, B.T. (2006) Risk and protective factors predictive of adolescent
pregnancy: A longitudinal prospective study, Applied Developmental Science, 10: 188–199.
Ekert-Jaffe, O., Joshi, H., Lynch, K., Mougin, R., Rendall, M.S. and Shapiro, D. (2002) Fertility,
timing of births and socio-economic status in France and Britain: social policies and occupa-
tional polarization, Population (English edition), 57: 475–507.
Elliott, J. and Shepherd, P. (2006) Cohort profile: 1970 British Birth Cohort (BCS70), International
Journal of Epidemiology, 35: 836–843.
Ermisch, J. and Pevalin, D. (2003a) Does a ‘teen-birth’ have longer-term impacts on the mother?
Evidence from the 1970 cohort study, ISER Working Papers, Institute for Social and Economic
Research, University of Essex, Colchester.
Ermisch, J. and Pevalin, D. (2003b) Who has a child as a teenager?, ISER Working Papers, Institute
for Social and Economic Research, University of Essex, Colchester.
Feinstein, L., Lupton, R., Hammond, C., Mujtaba, T., Salter, E. and Sorhaindo, A. (2008) Public
Value of Social Housing: A Longitudinal Study of Relationships Between Housing and Life
Chances, Smith Institute, London.
Ferri, E. (1993) Life at 33: The Fifth Follow-up of the National Child Development Study, National
Children s Bureau, London.
Geronimus, A. (1997) Teenage childbearing and personal responsibility: an alternative view, Polit-
ical Science Quarterly, 112: 405–430.
Geronimus, A. (2003) Damned if you do: culture, identity, privilege, and teenage childbearing in
the United States, Social Science & Medicine, 57: 881–893.
Goodman, A., Kaplan, G. and Walker, I. (2004) Understanding the effects of early motherhood
in Britain: the effects on mothers, IFS Working Paper 04/18, Institute for Fiscal Studies,
London.
Greene, M.E. and Biddlecom, A.E. (2000) Absent and problematic men: demographic accounts of
male reproductive histories, Population and Development Review, 26: 81–115.
Hadfield, L., Rudoe, N. and Sanderson-Mann, J. (2007) Motherhood, choice and the British media:
a time to reflect, Gender and Education, 19: 255–263.
Harden, A., Brunton, G., Fletcher, A., Oakley, A., Burchett, H. and Backhans, M. (2006) Young
people, pregnancy and social exclusion: A systematic synthesis of research evidence to iden-
tify effective, appropriate and promising approaches for prevention and support, The Evidence
for Policy and Practice Information and Co-ordinating Centre (EPPI-Centre), Social Science
Research Unit, Institute of Education, University of London, London.
Harrell, F.E.J. (2001) Regression Modeling Strategies; With Applications to Linear Models, Logis-
tic Regression and Survival Analysis, Springer, New York.
Hobcraft, J. and Kiernan, K. (2001) Childhood poverty, early motherhood and adult exclusion,
British Journal of Sociology, 52: 495–517.
Hoem, B. (2000) Entry into motherhood in Sweden: the influence of economic factors on the
rise and fall in fertility, 1986–1997, Demographic Research, 2, online journal available at
http://www.demographic-research.org/volumes/vol2/4/2-4.pdf
Hosmer, D.W.J. and Lemeshow, S. (2000) Applied Logistic Regression, Hoboken, New Jersey.
Hotz V.J., McElroy-Williams, S. and Sanders, S.G. (2004) Teenage childbearing and its lifecy-
cle consequences: exploiting a natural experiment, Working Paper, Department of Economics,
University of California (UCLA), Los Angeles.
Imamura, M., Tucker, J., Hannaford, P., Oliveira da Silva, M., Astin, M., Wyness, L.,
Bloemenkamp, K.W.M., Jahn, A., Karro, H., Olsen, J. and Temmermen, M. (2007) Factors
associated with teenage pregnancy in the European Union countries: a systematic review, Euro-
pean Journal of Public Health, 17: 630–636.
Jaffee, S.R., Caspi, A., Moffitt, T.M., Taylor, A. and Dickson, N. (2001) Predicting early fatherhood
and whether young fathers live with their children: prospective findings and policy reconsider-
ations, Journal of Child Psychology and Psychiatry, 42: 803–815.
Joshi, H. (2007) Social Polarisation in reproduction. From Boom to Bust: Fertility, Ageing and
Demographic Change, CentreForum, London.
Kendall, C., Afable-Munsuz, A., Speizer, I., Avery, A., Schmidt, N. and Santelli, J. (2005) Under-
standing pregnancy in a population of inner city women in New Orleans – results of qualitative
research, Social Science & Medicine, 60: 297–311.
Kiernan, K. and Hobcraft, J. (1997) Parental divorce during childhood: age at First intercourse,
partnership and parenthood, Population Studies, 51: 41–55.
Kneale, D. (2008) Pathways to parenthood: exploring the influence of context as a predictor of early
parenthood, Working Paper, Centre for Longitudinal Studies, Institute of Education, University
of London, London.
Kneale, D. and Joshi, H. (2008) Postponement and childlessness – evidence from two British
cohorts, Demographic Research, 19: 1935–1968.
Office for National Statistics (2007) Birth statistics: review of the Registrar General on births
and patterns of family building in England and Wales, 2006, Series FM1, Office for National
Statistics, Newport.
Plewis, I., Calderwood, L., Hawkes, D. and Nathan, G. (2004) National Child Development Study
and 1970 British Cohort Study Technical Report: Changes in the NCDS and BCS70 Popula-
tions and Samples over Time, Centre for Longitudinal Studies Technical Reports, Centre for
Longitudinal Studies, Institute of Education, University of London, London.
Pollock, G. (2008) Youth transitions: debates over the social context of becoming an adult, Sociol-
ogy Compass, 2: 467–484.
Power, C. and Elliott, J. (2006) Cohort profile: 1958 British birth cohort (National Child Develop-
ment Study), International Journal of Epidemiology, 35: 34–41.
Rendall, M.S., Clarke, L., Peters, H.E., Ranjit, N. and Verropoulou, G. (1999) Incomplete reporting
of men’s fertility in the United States and Britain: a research note, Demography, 36: 135–144.
Robson, K. and Berthoud, R. (2006) Age at first birth and disadvantage among ethnic groups in
Britain, Ethnic and Racial Studies, 29: 153–172.
Rutter, M. (1967) A children’s behaviour questionnaire for completion by children: preliminary
findings, Journal of Child Psychology and Psychiatry, 8: 1–11.
Schoon, I., Bynner, J., Joshi, H., Parsons, S., Wiggins, R. and Sacker, A. (2002) The influence of
context, timing and duration of risk experiences for the passage from childhood to midadult-
hood, Child Development, 73: 1486–1504.
Simpson, R. (2006) Childbearing on hold: a literature review, Working Paper 1, Cen-
tre for Research on Families and Relationships, University of Edinburgh, Edin-
burgh. www.uptap.net/reports%20and%20papers/05UPTAP%20Working%20Paper%201
R.Simpson.doc
Singh, S., Darroch, J.E. and Frost, J.J. (2001) Socioeconomic disadvantage and adolescent
women s sexual and reproductive behaviour: the case of five developed countries, Family Plan-
ning Perspectives, 33: 251–258+259.
Social Exclusion Unit (1999) Teenage Pregnancy, Social Exclusion Unit, London.
Van de Kaa, D.J. (1987) Europe s second demographic transition, Population Bulletin, 42: 1–59.
Chapter 6
Currently Cohabiting: Relationship Attitudes,
Expectations and Outcomes
Ernestina Coast
Introduction
The rise in cohabitation – pre-marital, non-marital and post-marital – represents

one of the most significant changes in union formation patterns in many developed
economies. The importance of cohabitation, and the public debates it generates,
are reflected in the media attention it has received. In 2006, there were 2.3 mil-
lion cohabiting couple families in the UK (ONS, 2007). 1 The increase in cohab-
itation has occurred alongside other, related, major demographic shifts, including:
rising levels of divorce; delay in entry into marriage and childbearing; and a rise
in the proportion of births taking place outside marriage. These are all charac-
teristic of the second demographic transition (Van de Kaa, 1987; Lesthaeghe and
Surkyn, 2004b), although rising levels of cohabitation in the UK have only par-
tially offset declining marriage rates (Berrington and Diamond, 2000). Even within
Europe, divergent trends in the timing, duration, type and composition of cohab-
iting unions have been identified (Liefbroer and Dourleijn, 2006; Kiernan, 2001;
2004; Prinz, 1995). Theorising about cohabitation encompasses a broad range of
perspectives, from notions of selfish individualism and breakdown of the family
(Morgan, 2000) to those of the democratic, consensual and “pure” relationship
(Giddens, 1992; Beck-Gernsheim, 2002).
Cohabitation may be narrowly defined as “an intimate sexual union between
two unmarried partners who share the same living quarter for a sustained period
of time” (Bachrach et al., 2000), assuming a defined duration (Mynarska and
Bernardi, 2007). Typologies of cohabitation continue to evolve (Haskey, 2001;
Martin and Thery, 2001; Casper and Bianchi, 2002), reflecting the changing
nature of living arrangements in general and cohabitation in particular. As both a
E. Coast (B)
Department of Social Policy, London School of Economics, Houghton Street,
London, WC2A 2AE, UK
e-mail: E.Coast@lse.ac.uk
1 Thereare no annual official estimates of the cohabiting population of England and Wales, unlike
legal marital status, and trends tend to be derived from surveys such as the Labour Force Survey
and General Household Survey (GAD).

106 E. Coast
demographic process and event, cohabitation is fuzzy (Knab, 2005), elusive (Teitler
and Reichman, 2001) and heterogeneous (Oppenheimer, 2003). Union formation
in general, and cohabitation in particular, are characterised by increasing number
and complexity, with the duration of cohabiting unions appearing to be lengthening
(Haskey, 2001).
Theorists seeking to explain the rise in cohabitation incorporate a wide range
of explanatory perspectives, including: increased secularization (Lesthaeghe, 2002;
Thornton et al., 1992; Lehrer, 2004); increased female labour force participation;
shifts in the meaning of marriage (Allan and Crow, 2001), including a decline in
its socio-cultural function (Alders and Manting, 2001); risk reduction (Mulder and
Manting, 1994; Galland, 1997); a decline in the cultural importance of kin; and the
separation of sex and reproduction.
Cohabitation may now be considered normative in the UK, evidenced by survey
and opinion poll data. Such attitudinal data can contribute to the body of evidence
about prevailing social norms (and stigma) and associated behaviour. Attitudinal
data about cohabitation provide one strand of evidence about the acceptability of
cohabitation as a social institution, and contribute to the substantive demographic
evidence about the role of cohabitation in contemporary societies. Responses to
questions about attitudes to cohabitation reveal the extent to which individuals
have internalised norms about appropriate and ‘normal’ behaviour with respect to
union formation (Oropesa, 1996). In 1981, a special edition of the journal Alterna-
tive Lifestyles dealt with cohabitation as a new form of living arrangement. Today,
cohabitation has moved from being a ‘deviant’ or ‘alternative’ lifestyle choice
to one that is normative (de Vaus et al., 2005), both before and after marriage
(Bumpass et al., 1991, 1995).
Context
Normative Attitudes
Attitudes are inherently subjective and virtually impossible to verify. When inter-
preting attitudinal data generated by surveys, it is important to note that respondents
have to create judgements quickly in response to the question asked, often in rela-
tion to some implicit standard, even if the judgements are themselves rooted in a
firmly held view (Tourangeau et al., 2000). Whilst all survey questions are context-
dependent, evidence shows that attitudinal questions are particularly at risk of this
effect (Schuman and Presser, 1996). There are two broad types of attitudinal survey
data: normative and individual. Normative attitudinal responses, such as those col-
lected in opinion polls, allow an individual to distance themselves from their own
circumstances. Individual attitudinal responses are, theoretically, grounded in real-
ity. Norms and values relating to union formation are dynamic and respond to the
interaction between individual experiences and social responses (Bachrach et al.,
2000) and both contribute to, and arise from, changes in society (Heuveline and
6 Currently Cohabiting 107
Timberlake, 2004). Attitudinal surveys are used extensively in policy debates, for
example, recent debates about the legal position of cohabiting relationship in Britain
have incorporated attitudinal information as part of their corpus of evidence for legal
change (Dey and Wasoff, 2007).
Changes in normative attitudes towards cohabitation are poorly represented
before the final quarter of the twentieth century, mirroring the paucity of substantive
data on the prevalence of non-marital cohabitation (Gillis, 2004; Murphy, 2000a, b;
Kiernan, 2004). Globally, attitudes toward pre- and non-marital cohabitation have
become more ambivalent and less unaccepting of non-traditional living arrange-
ments in general, and cohabitation in particular (Thornton, 1989). For example,
Thornton’s research identifies a clear trajectory of changing attitudes towards cohab-
itation in the US, with rapid changes in the 1960s and 1970s, slowing down in the
1980s. Normative differences in attitudes towards cohabitation have been studied
in a variety of comparative settings, particularly in the US (Carter, 1993; Oropesa,
1996; Thornton, 1989; Sweet and Bumpass, 1992; Thornton and Young-DeMarco,
2001; Nock, 1998; Heuveline and Timberlake, 2004; Axinn and Thornton, 2000;
Thornton, 1995), and elsewhere such as Sweden (Bernhardt, 2004; Trost, 1978),
Europe (Kiernan, 2004), the UK (Haskey, 2001; Barlow et al., 2001), and Poland
(Kwak, 1996; Mynarska and Bernardi, 2007). An increase in the acceptability of
cohabitation can reasonably be interpreted as evidence for weakening of the social
norms surrounding marriage, referred to variously as the deinstitutionalisation of
marriage (Cherlin, 2004), démariage (Théry, 1993) and the disestablishment of mar-
riage (Coontz, 2004, quoting Cott).
Responses to normative questions are grounded in a specific time and context.
Because cohabitation (and other forms of intimate relationship) are dynamic – a
moving target, responses to questions about the acceptability of cohabitation posed
in the 1980s potentially have different meanings than responses to questions posed
in the twenty-first century, even if the question wording and response categories
are exactly the same. In Britain, two key sources of population-level attitudi-
nal data about cohabitation are the British Household Panel Survey (BHPS) and
British Social Attitudes Survey (BSA). The data reported here update and expand
upon Haskey’s (2001) review of population-level attitudes towards cohabitation in
Britain.
The BHPS has asked a series of repeated self-completion questions about atti-
tudes towards cohabitation using Likert-scale responses. It is important to note that
the phrasing of the self-completion questions changed at Wave 8 (1998). Previous
waves (1992, 1994, 1996) used the statement “Living together outside of marriage
is always wrong”. Subsequent waves (1998–2004) used the statement “It is alright
for people to live together even if they have no interest in considering marriage”.
The BHPS also includes a separate youth questionnaire for all household members
aged 11–15 years (inclusive), incorporating the repeated statement response “Liv-
ing together outside of marriage is always wrong”. Interestingly, whilst this phras-
ing was changed for adult BHPS respondents, it has remained constant for youth
respondents.
108 E. Coast
The annual cross-sectional BSA survey has included a combination of repeat

and ad hoc2 questions about attitudes to cohabitation (Haskey, 2001; Barlow, 2004).
This review focuses on repeat elements, namely attitude responses to statements in
1989, 1994, 2000 and 2002 (Do you agree or disagree? It is a good idea for a couple
who intend to get married to live together first” “Do you agree or disagree? It is all
right for a couple to live together without intending to get married.” “Do you agree
or disagree? People who want children ought to get married”). Table 6.1 summarises
normative data relating to cohabitation in the BHPS (1992–2004). More than two
thirds of respondents have reported agreement with the statement “It is alright for
people to live together even if they have no interest in considering marriage” in each
of four successive waves.
Disaggregating responses to statements about cohabitation by birth cohort, a
clear generational pattern emerges, with older cohorts much less likely to approve
of non-marital cohabitation relative to younger cohorts. This is mirrored by trends
in reported ever-cohabitation by birth cohort. Less than 3% of respondents born
Table 6.1 Percentage distribution of respondents’ attitudes to cohabitation in general, 1992–2004
“Living together outside of “It is alright for people to live together

marriage is always wrong” even if they have no interest in
considering marriage”
1992 1994 1996 1998 2000 2002 2004
Strongly agree/Agree 16.7 14.7 14.1 66.4 69.7 68.4 69.4

Neither agree nor disagree 30 28.9 28.5 21.4 19.5 20.3 20
Strong disagree/ Disagree 53.3 56.4 57.4 12.2 10.8 11.3 10.6
N 9,284 8,940 9,027 10,427 14,799 15,215 14,341
Note the phrasing of the self-completion questions changed at Wave 8 (1998). Previous waves
(2, 4, 6) used the statement “Living together outside of marriage is always wrong”. Subsequent
waves (8, 10, 12, 14) used the statement “It is alright for people to live together even if they have
no interest in considering marriage”. n = valid cases, excluding missing or don’t know responses.
Source: BHPS 1992–2004 documentation and questionnaires: http://www.iser.essex.ac.uk/ulsc/b
hps/doc/volb/indexes/subjcat20.php#Values,%20Opinions%20and%20AttitudesAccessed30/06/08.
2 BSA ad hoc questions on attitudes towards cohabitation are as follows: 1986 “Do you agree or
disagree? As a society we ought to do more to safeguard the institution of marriage”. 1986 “Do
you agree or disagree? Most people nowadays take marriage too lightly”. 1989 “Do you agree or
disagree? Personal freedom is more important than the companionship of marriage”. 1989 “If you
were advising a young (wo)man, which of the following ways would you recommend? Live alone
with no partner /Live with a partner and not marry / Live with a partner and then marry/ Marry
first”. 1989 and 1994 “Do you agree or disagree? The main advantage of marriage is that is gives
financial security”. 1994 “Imagine an unmarried couple who decide to have a child, but do not
marry? What would your general opinion be?”. 2000 “Many people who live together without
getting married are just scared of commitment”. 2000 “There is no point getting married - it’s
only a piece of paper”.
Agree statement male

Percentage (strongly) agree response to Agree statement female Ever cohabited male Ever cohabited female
the statement Òit is alright for people to 100 70
live together even if they have no
interest in considering marriage 90
60
Percentage ever cohabited

80
70 50
60
40
50
30
40
30 20
20
10
10
0 0
1920 1930 1940 1950 1960 1970
Birth cohort
Fig. 6.1 Distribution of attitudes towards, and experience of, cohabitation, by birth cohort and
sex, 1920–1970
(Source: BHPS, 2004.)
in the 1920s reported ever having cohabited, compared with 57% of respondents
born in the 1970s. Men appear to have slightly more accepting attitudes towards
cohabitation, although this differential is negligible for more recent birth cohorts
(Fig. 6.1).
Individuals who have ever-cohabited are significantly (p<0.000) more likely to
report approving attitudes towards cohabitation, with just 1.7% of ever-cohabiting
respondents disagreeing with the statement “It is alright for people to live together
even if they have no interest in considering marriage”. This significant relation-
ship holds for all birth cohorts. It is possible to examine whether an individ-
ual’s attitude towards cohabitation in general changes over the six year inter-
val between the first (1998) and most recent (2004) waves including attitudinal
statements on non-marital cohabitation. Normative attitudes reported in the BHPS
are relatively stable. Nearly three quarters (74%) of respondents report the same
broad response in both 1998 and 2004. Of those that change their broad response
category over the period, the majority shift towards a more accepting attitude
over time.
Successive generations tend to have less traditional attitudes when compared
with preceding generations, a function of both generation succession and intra-
generational change (Scott et al., 1996). Adolescents’ attitudes provide insight into
the probable trajectory of normative attitudes and behaviours in the near future.
The attitudes of adolescents are important for determining future choices (Burt and
Scott, 2002; Manning et al., 2007 ), with young adults who approve of cohabita-
tion more likely to enter into a cohabiting relationship (Axinn and Thornton, 1993).
Successive BHPS waves (1994, 1999, 2000, 2001, 2005) asked young people aged
110 E. Coast
Table 6.2 Percentage distribution of youths aged 11–15 years (inclusive) response to the question
statement “Living together outside of marriage is always wrong”
Living together outside of marriage is always wrong

1994 1999 2000 2001 2005
Strongly agree/Agree 19.0 12.8 11.9 10.6 13.3

Neither agree nor disagree 21.2 27.9 26.7 23.5 30.8
Strong disagree/Disagree 59.8 59.3 61.4 65.9 55.9
n 759 929 1,409 1,404 1,401
n = valid cases, excluding missing or don’t know responses.
Source: BHPS 1994–2005 documentation and questionnaires: http://www.iser.essex.ac.uk/ulsc/
bhps/doc/volb/indexes/subjcat20.php#Values,%20Opinions%20and%20Attitudes.
11–15 years their attitude toward the statement “Living together outside of mar-
riage is always wrong”. Treating the data as cross-sectional for descriptive pur-
poses, the broad pattern appears to be one of increasing ambivalence, with nearly
one third of respondents neither agreeing nor disagreeing with the statement in 2005
(Table 6.2).
When using panel data to examine normative attitudes, it is useful to try to
disentangle whether observed changes in attitudes develop because the observed
individuals have adopted new attitudes or because new individuals with differ-
ent attitudes have entered the population. For example, the five year gap between
the 1994 and 1999 rounds of the youth questions on cohabitation meant that
very few of those interviewed in 1994 would have still been eligible for inter-
view in 1999. By contrast, repeat of the questions in successive years (1999,
2000, 2001) would have repeatedly captured a proportion of the population who
remained within the 11–15 age group. The BSA has asked a set of questions
about attitudes towards cohabitation in subsequent survey years (1994, 1998,
2000, 2002). The proportion of individuals expressing negative views about
cohabitation, and its relation to marriage, has declined across all age groups
(Fig. 6.2).
Questions about attitudes towards cohabitation are just one element of the bat-
tery of attitudinal questions contained in surveys such as the BSA. For example, the
BSA has asked questions annually about non-marital (pre- and extra-) sexual rela-
tionships since 1983. It is important to note that attitudes of increased acceptance of
cohabitation have changed more rapidly than attitudes towards other aspects of inti-
mate relationships such as extra-marital sex and same-sex relationships. As such,
cohabitation has emerged as an aspect of intimate relationships that has come to
be regarded differently (perhaps separately?) from other indicators of sexual free-
dom (Murphy, 2000b; Reynolds and Mansfield, 1999). Acceptance of cohabitation
is likely to increase in the future, a function of the social processes of cohort replace-
ment, socialisation and social diffusion (Seltzer, 2004).
50
Percentage of respondents
45
40
35
30
25
20
15
10
5
0
1994
2002
1994
2002
1994
2002
1994
2002
1994
2002
1994
2002
1994
2002
1994
2002
1994
2002
18–34 35–54 55+ 18–34 35–54 55+ 18–34 35–54 55+
It is all right for a It is a good idea People who want
couple to live together for a couple who intend children ought to get
without intending to to get married to live married
get married together first
Fig. 6.2 Percentage distribution, by age group,3 of respondents who disagree, or strongly disagree,
with statements about cohabitation and marriage, 1994 and 2002
(Source: BSA.)
Individual Attitudes, Expectations and Outcomes
In the USA, a limited number of studies have examined the effects of cohabitors’
own assessments of their relationship on union outcomes. Brown (2000) combines
both relationship assessments (positive versus negative) and expectations. Research
elsewhere has considered the relationship between relationship expectations and
outcomes. For example, Manning and Smock (1995) found that cohabiting couples
that express an intention to marry are four times more likely to marry compared
with couples with no reported plans to marry. Reports of plans or expectations to
marry by cohabiters can be interpreted as indicative of cohabiting unions repre-
senting a transitional state leading to marriage. Relationship expectations cannot be
used as proxy indicators of relationship ‘quality’. For example, an expectation of
relationship transition to marriage might be an expression of a perceived absence of
alternatives to the current cohabiting relationship. Similarly, an expectation of split-
ting up may be an expectation based on externalities such as forthcoming university
attendance in another part of the country.
Much research compares cohabitation to marriage, and compared to married
couples, cohabiting couples differ in several distinct ways (Bachrach et al., 2000)
including higher rates of union instability (Waite and Gallagher, 2000; Ermisch and
Francesconi, 2000; Bouchard, 2006). It is important to note, however, that much
of the earlier research into the dissolution of cohabiting versus married partner-
ships used data on unions from the 1970s and early 1980s when cohabitation was
3 Age at time of interview.

112 E. Coast
much rarer (de Vaus et al., 2005). The influence of cohabitation has been examined
on a wide range of outcomes, and is associated with: relationship dissatisfaction
(DeMaris, 1984); higher levels of conflict and violence (Thomson and Collela, 1992;
Forste, 2002; Kenney and McLanahan, 2006); lower quality of partner communica-
tion (Cohan and Kleinbaum, 2002); lower levels of sexual exclusivity (Forste and
Tanfer, 1996), greater dependency on family of origin (Rindfuss and VandenHeuvel,
1990); and lower relationship quality (Brown and Booth, 1996).
Relationship expectation reports are cross-sectional, and it might be that an indi-
vidual entered into a cohabiting relationship with no expectations of marriage, but
that these expectations changed over time. The absence of expectations to marry
can represent one of three positions: firstly, an ideological position that opposes
marriage; secondly, an assessment that their current partner is not marriage mate-
rial, but an absence of an ideological opposition to marriage per se; thirdly, they
have yet to transition to thinking about marriage. The purpose of looking at rela-
tionship expectations is to throw some light on whether cohabitation represents an
alternative to marriage, or an integral component of the transition to marriage. For
example, older cohabitors tend to be more likely to report their relationship as an
alternative to marriage, whereas younger cohabitors are more likely to report cohab-
itation as a precursor to marriage (King and Scott, 2005). It is important to anal-
yse gendered relationship expectations and attitudes. Considerable research into the
gendered aspects of marriage has revealed ‘his’ and ‘her’ marriages, first identified
by Bernard and Bernard (1982) and subsequently Fowers (1991), and it is reason-
able to hypothesise that there are ‘his’ and ‘her’ cohabitations.
Analyses
Data
This research uses data from the BHPS to analyse individuals’ relationship expecta-
tions and subsequent reported relationship behaviour (University of Essex, 2006). It
deals with the relationship intentions of those individuals who report a non-marital
cohabiting partner. How do cohabiting relationship expectations differ by age, sex,
previous relationship history and parenthood? For people in cohabiting relation-
ships, how do attitudes towards cohabitation differ by age, sex, previous relation-
ship history and parenthood? Do individuals achieve their relationship expectations?
How are cohabiting couples’ relationship expectations associated with relationship
outcomes (marriage, separation, continued cohabitation)?
Begun in 1991, the BHPS surveys approximately 5,000 households annually. In
the eighth wave, in 1998, and again in the thirteenth wave, in 2003, individuals
aged 16 and above who were in cohabiting relationships were asked about their
expectations of this cohabiting relationship. They were shown a card with a range
of responses and asked to “read out the number of the statement which you feel
applies most closely to your current relationship”. The responses included: “Plan-
ning to marry”, “Probably get married”, “Just live together”, “No thought to the
future”, “Don’t know” and “Other”. A supplementary question was asked of those
respondents who replied “Don’t know” or “Just live together”. The supplementary
question also used a showcard, and asked for a response to the statement “how
likely it is that you will ever get married (or remarried) to anyone in the future?”.
The responses included: “Very likely”, “Likely”, “Unlikely”, “Very unlikely” and
“Don’t know”.
Cohabiting respondents were asked a series of questions about their perceptions
about cohabitation in general. It is important to note that this series of questions did
not explicitly ask respondents to reflect upon their own current cohabiting relation-
ship, but the questions did explicitly compare cohabiting relationships to marriage,
rather than to any other form of union. The questions were: “Do you think there are
any advantages in living as a couple, rather than being married?” and “Do you think
there are any disadvantages in living as a couple, rather than being married?”. If a
respondent answered yes to either of these questions, they were prompted for open
ended responses (up to two mentions) with the question “What do you think are the
(dis)advantages of living as a couple?”
Also in the eighth wave, in 1998, and again in the thirteenth wave, in 2003, indi-
viduals aged 16 and above were asked “Do you have a steady relationship with a
male or female friend whom you think of as your ‘partner’, even though you are
not living together?” Respondents that reported such a partner were then asked their
intentions about this relationship, based on show card responses to the question
“Please look at this card and read out the number of the statement you feel applies
most closely to this relationship?”, with responses of “Expect to marry”, “Expect to
cohabit”, “No plans to marry or cohabit” and “Don’t know”. For those individuals
who reported a partner, but did not report an expectation of marriage or cohab-
itation with this partner, a supplementary showcard response question was asked
“Can you please look at this card and tell me how likely it is that you will ever
get married or remarried to anyone in the future?”, with responses of “Very likely”,
“Likely” “Unlikely”, “Very unlikely” and “Don’t know”. Because the BHPS only
collects data from coresidential members of a household, data and analyses on non-
co-residential partners are restricted to those individuals who are members of the
BHPS sample, and not their partners.
Descriptive Overview
Cohabitation is heterogeneous, involving pre-, intra- and post-marital cohabiting

relationships (Table 6.3). For women and men born in the 1970s, 72 and 75%,
respectively, of first unions were cohabiting. The normative status of cohabitation
as a first type of union is underlined by examining the relatively small numbers of
individuals born in the 1980s and aged 16 and over included in the BHPS. Of those
members of this most recent cohort who have entered live-in unions (n = 470), 91%
report cohabitation as the first type of union, underlining the primacy of cohabitation
as first union (Berthoud, 2000).
114 E. Coast
Table 6.3 Percentage distribution of respondents reporting a cohabiting relationship, by sex, 1998
and 2003
1998 2003
n = 1,187 n = 1,511
Current legal marital status

– Married 1.2 1.9 1.1 0.5
– Separated 3.2 3.2 3.7 1.9
– Divorced 24.0 25.6 22.9 25.8
– Widowed 1.4 2.3 1.2 2.3
– Never married 70.1 67.0 71.0 69.5
Parent 32.0 40.8 40.5 50.8
Length of cohabiting
relationship at interview
– < 6 months 15.6 16.4 6.9 8.8
– 6–12 months 13.1 13.8 10.7 9.5
– 1–2 years 18.2 16.4 17.2 17.9
– 2–5 years 31.3 29.4 29.8 30.4
– > 5 years 21.8 23.9 35.4 33.6
Source: BHPS, 1998 and 2003, author’s analysis.
In terms of the characteristics of individual cohabiting couples, more than half

(55.1%) of all cohabiting relationships in 2003 where neither partner had been pre-
viously married, involved both partners born in the 1970s.
Cohabiting Relationship: Attitudes
Cohabiting relationship attitudes and expectations were collected in both 1998 and
2003. When examining whether relationship expectations are achieved, data are
used from the 1998 wave forwards to the most recent interview for each individ-
ual. The short time span between the 2003 wave of relationship expectation data
and the most recent published wave of the BHPS (2005) precludes detailed analysis
of relationship outcomes from the 2003 wave forwards.
An individual can report both advantages and disadvantages of cohabitation
when compared to marriage, and the two are not mutually exclusive. In 1998 and
2003 the majority of cohabiting respondents reported neither an advantage nor a dis-
advantage (47 and 55% respectively) (Table 6.4). In 2003 less than one third of indi-
viduals in cohabiting relationships reported that there was an advantage to living in a
cohabiting relationship when compared to marriage. Responses from never-married
individuals are based on perceptions about marriage, rather than direct experience of
it. Overall, there is no significant relationship between sex and whether an individ-
ual reports an advantage to cohabitation compared to marriage. However, respon-
dents who are parents are significantly (1998 p<0.005; 2003 p<0.000) less likely
to report advantages of cohabitation compared to non-parents. Examining in detail
Table 6.4 Percentage distribution of attitudes about cohabitation, currently cohabiting respon-
dents, 1998 and 2003
1998 2003
n=1,115 n=1,514
Advantages in living as a couple 40.0 32.0

First mentioned advantage1
– trial marriage 30.7 23.6
– no legal ties 29.8 24.5
– improves relationship 5.2 3.6
– previous bad marriage 1.6 2.7
– personal independence 10.0 10.9
– financial advantage 16.1 22.2
– companionship 2.0 3.1
– prefer cohabitation 1.4 1.3
– other 3.2 8.2
Disadvantages in living as couple 26.7 23.6
First mentioned disadvantage2
– financial insecurity 39.0 30.4
– no legal status 16.6 32.1
– effects on children 5.4 6.2
– lack of commitment 15.6 9.6
– social stigma 16.3 11.3
– other 7.1 10.4
1
Second mentioned advantages were collected in both 1998 and 2003,
but have not been included in analyses here due to the relatively small
numbers (n = 44, n = 33, respectively) reporting a second advantage.
2
Second mentioned disadvantages were collected in both 1998 and
2003, but have not been included in analyses here due to the relatively
small numbers (n = 16, n = 44, respectively) reporting a second disad-
vantage. Source: BHPS, 1998 and 2003, author’s analysis.
the four most commonly reported advantages of cohabitation (trial marriage, no

legal ties, personal independence and financial advantage), there are no significant
differences by sex. Parenthood status is significantly related (1998 p<0.001; 2003
p<0.005), with non-parents more likely to report trial marriage, and parents more
likely to report personal independence and the absence of legal ties as advantages
of cohabitation.
Approximately one quarter of respondents report disadvantages in living as a
couple in both 1998 and 2003, with women significantly more likely to report dis-
advantages compared to men if they had a previous live-in relationship (p<0.000)
or were a parent (p<0.050). For the subset of individuals whose cohabiting rela-
tionships extended across the 1998 and 2003 interviews (n = 144), it is possible to
examine the consistency of responses over time. Of those individuals reporting atti-
tudes on the same cohabiting union in 1998 and 2003 (n = 132), overall attitudes are
fairly consistent, reporting the same response to whether there are advantages or dis-
advantages (62.1 and 68.9%) to cohabitation. This suggests that those individuals in
long duration cohabiting relationships have well-established attitudes towards their
116 E. Coast
Table 6.5 Percentage distribution of responses to the statement “How likely is it that you will ever
get married to anyone in the future?”, by currently cohabiting, never married respondents with no
plans to marry their current partner, by sex, 1998 and 2003
1998 2003
n = 268 n = 401
Don’t know 11.6 8.6 10.2 6.8

Very likely 4.7 5.8 3.1 3.4
Likely 24.0 28.8 18.9 23.9
Unlikely 25.6 38.8 40.8 42.9
Very unlikely 34.1 18.0 27.0 22.9
union. Substantial proportions of never-married, currently cohabiting respondents

with no expectation of marriage for the current cohabiting relationship, report that
they are unlikely or very unlikely ever to marry, with 67.8 and 65.8% of men and
women, respectively, reporting this expectation (Table 6.5).
The percentage of those respondents who reported no plans to marry in 1998, and
thought it was (very) unlikely they would ever marry, did actually go on to marry
(18.5% split up and 71.4% were still cohabiting at their most recent interview).
Cohabiting Relationships: Expectations
If cohabitation is part of the marriage process, then one might reasonably expect
individuals to respond that they have plans to marry the longer they have cohab-
ited. For cohabiting individuals interviewed in 2003, the relationship between the
duration of the cohabiting relationship is significantly (p<0.000) associated with
relationship intentions (Table 6.6).
The majority of individuals in what might be described as long-term cohabiting
relationships do not report an expectation of marriage, but of continued cohabi-
Table 6.6 Percentage distribution of future relationship expectations, by duration of current cohab-
iting relationship, 2003
Future of current cohabiting relationship

Plan to marry Probably marry Live together
Duration of current < 1 year 30.5 38.0 31.6

cohabiting relationship 1–2 years 29.9 44.4 25.7
2–5 years 19.8 48.5 31.7
> 5 years 9.2 33.4 57.4
n = 1,1015 respondents
Source: BHPS, 2003, author’s analysis.
Table 6.7 Distribution of expectations by prior relationship history, 1993 and 2003
Future of current 1998 2003

cohabiting (n = 1,007) (n = 1,343)
relationship
No previous live-in Prior live-in No previous live-in Prior live-in
relationship relationship relationship relationship
Planning to marry 24.7 13.3 22.7 16.9

Probably marry 46.8 37.6 47.2 33.7
Live together 28.5 49.0 30.1 49.4
tation. The BHPS does not collect information on whether a couple has become
engaged – in and of itself not a formal or binding event – but it is reasonable to infer
that individuals with relatively short-lived cohabiting relationships have moved in
because a marriage is already planned. Individuals who had a prior live-in relation-
ship (whether married or cohabiting) are significantly (p < 0.000 for both 1998
and 2003) more likely to report an intention to continue cohabiting compared with
individuals who have not had a prior live-in relationship (Table 6.7).
Cohabiting Relationships: Outcomes and Expectations

Table 6.8 shows the distribution of outcomes of cohabiting relationships identified in
1998. For never-married, childless respondents interviewed in 1998, the subsequent
birth of a child within the relationship is significantly (p<0.050) associated with the
relationship outcome, with subsequent parents more likely to continue to cohabit
and less likely to marry compared to non-parents.
What proportion of individuals achieve their relationship expectations? Based
on responses to questions about cohabiting relationships in 1998, it is possible to
examine the outcome of those relationships to the most recent interview (Table 6.9).
For those respondents that reported a ‘definite’ expectation (plan to marry/ prob-
ably marry/continue to cohabit), there is a highly significant (p<0.000) relationship
between expectation and outcome, for both men and women and for both parents
Table 6.8 Percentage distribution of relationship outcomes, for cohabiting unions identified in
1998.
Subsequent outcome
Split up Marry Continue to cohabit
All 16.6 30.3 53.1

Never married 17.3 31.2 51.5
Ever-married 15.1 28.4 56.5

118 E. Coast
Table 6.9 Percentage distribution of outcome of cohabiting relationships by relationship expecta-

tions expressed in 1998
Future of current relationship

Outcome to Plan to Probably Live No thought Do not
date marry marry together to future know
Split up 0.9 7.0 6.6 1.3 0.1

Marry 10.7 13.6 4.8 0.6 0.1
Continue to 4.2 20.9 23.9 3.7 1.5
cohabit
and non-parents at the time of interview. More than two thirds (67.9%) of those
individuals who reported that they planned to marry their cohabiting partner then
went on to marry that partner.
In order to examine concordance and discordance of relationship expectations
between men and women, we select couples where both partners provided full
responses to questions about the advantages and disadvantages of cohabitation, and
their expectations of the current cohabiting relationship (Table 6.10). Analyses here
are restricted to those individuals reporting on cohabiting unions which represented
their first ever live-in relationship. Because analyses are based on fully respond-
ing couples, the responses may be biased for homogeneity of response (Berrington,
2004). It is possible to identify whether anyone else was present during the BHPS
interview, but interviewers report very low levels of influence of third parties when
they are present during interview. 4
Levels of concordance (either both report “Yes” or both report “No”) within cou-
ples are high, with most concordance for “No” responses to questions about dis-
advantages and advantages of cohabitation when compared with marriage. Where
both partners report an advantage of cohabitation over marriage, the most common
concordant response is as a trial marriage, in both 1998 and 2003 (32.4 and 26.5%
Table 6.10 Percentage distribution of couple concordance on attitudes towards cohabitation
1998 2003
n=168 couples n=231 couples
Couple concordant % Couple concordant %
Advantages 65.4 64.9

Disadvantages 63.9 74.0
4 For example, in 1998, of 187 interviews of currently cohabiting couples, 108 (58 per cent) record
a third party as being present. 96 of these 108 interviews (89 per cent) are coded as no influence
exerted by the third party.
of couples). Levels of agreement within couples about specific disadvantages of

cohabitation are much lower, although financial insecurity is the most commonly
mentioned where both partners report a disadvantage.
In terms of future expectations about their current cohabiting union, there are
high levels of concordance within couples. Of course, concordance does not equal
achievement of these desires, concordant couples may still be disappointed in the
future.
For those cohabiting couples interviewed in 1998, it is possible to examine their
relationship outcomes by the date of their last interview (Table 6.11). 81.5% of those
couples who agreed in 1998 that they planned to marry did go on to marry, whereas
only 39.5% of those couples who agreed they would probably get married went on
to convert their relationship to a marriage. Three fifths of couples who agreed in
1998 that they would continue to cohabit were still cohabiting at their most recent
interview wave in the BHPS.
Table 6.11 Couple relationship expectations, currently cohabiting couples, 1998 and 2003
1998 2003
n = 137 couples n = 196 couples
Women Women
Planning to Probably get Just live Planning to Probably get Just live
Men marry married together marry married together
Planning to 20.4 8.0 0.7 19.9 5.1 1.0

marry
Probably get 3.6 43.8 5.8 3.6 37.8 10.7
married
Just live 0 5.8 11.7 0 4.6 17.3
together
Discussion
The data reported here underline the heterogeneity of cohabitation, a heterogeneity

that raises challenges for researchers to make generalisations about the processes
that underlie it, the forms it takes, and the intentions that people report.
In this study the majority of cohabitors assert that they will marry their partner
(including both “Plan to marry” and “Probably marry”), in keeping with US analy-
ses (Bumpass and Sweet, 1989; Carlson et al., 2004; Manning and Smock, 2002).
Such responses would imply that cohabitation is one element of the process of mar-
riage, and that cohabitation represents a considered step on the pathway to marriage.
However, what we cannot tell is whether these intentions to marry preceded becom-
ing a co-residential couple, or whether they emerged as a result of having co-resided.
Recent work suggests that many (if not most) cohabiting couples ‘slide’ rather than
120 E. Coast
‘decide’ into a co-residential cohabiting relationship (Stanley et al., 2006), echoing

findings from Lindsay’s (2000) work in Australia.
Current cohabitees who have a previous live-in relationship and are already par-
ents are more likely to report an expectation of cohabitation rather than marriage,
echoing work in the US (Bumpass et al., 1991). Smart and Stevens’ (2000) study of
cohabiting families in Britain reports that some cohabiting mothers prefer to con-
tinue cohabiting rather than marry a man whom they were uncertain they could rely
on for support or to enter into single parenthood. The reported advantages and disad-
vantages of cohabitation relative to marriage among current cohabitees in the BHPS
suggest that, at least for never-married respondents, assessing compatibility through
a ‘trial marriage’ is important. King and Scott’s (2005) work in the US using the
National Survey of Families and Households, reports that compatibility assessment
by younger cohabitors was a key reason for cohabitation.
The wording of survey questions such as those included in the BHPS tend to pose
statements about cohabitation relative to marriage. This standpoint reflects much
of the broader academic endeavour surrounding cohabitation, which has debated
whether cohabitation is a prelude to marriage, or whether it is an alternative to
marriage. A body of work has suggested, however, that a more productive line of
enquiry might be to view cohabitation as an alternative to being single (Rindfuss
and VandenHeuvel, 1990) and/or a progression of an intimate non-co-residential
relationship (Casper and Bianchi, 2002; McGinnis, 2003). It is also quite possible
that each rationale may operate at different points over time for an individual.
Datasets, including the BHPS, rarely collect information on engagements, which
affect the entering into, and dissolution of, cohabiting unions. If couples are cohab-
iting as a result of engagement with an intention to marry, then engagement-driven
cohabitation explains in part both the rise in cohabitation and delays in marriage.
Such compositional shifts (Oppenheimer, 2003) in cohabitation, further complicate
its study. One possible reason, rarely explored, for reported intentions not to convert
a cohabiting union into a marital union, is that of the costs of a wedding (Kravdal,
1999). Whilst a marriage in England and Wales costs approximately £100, the cost
of a wedding can run to tens of thousands of pounds, and for many people, the
marriage and the wedding are indivisible as processes (Otnes and Pleck, 2003).
Such cost-related concerns can become more sharply focused if one or both of the
cohabiting partners is a parent, notwithstanding other economic needs identified as
prerequisite to marriage (Gibson-Davis et al., 2005).
Future research needs to widen the pool of potential couples available to enter
into a co-residential union, whether cohabiting or married, and their relationship
intentions. ‘Living-apart-together’ (LAT) relationships, in which two partners
regard themselves as a couple but do not cohabit, have recently been recog-
nised in the social science literature (Levin and Trost, 1999; Bawin-Legros and
Gauthier, 2001; Karlsson and Borell, 2002; Borell and Karlsson, 2003; Milan
and Peters, 2003; Levin, 2004; de Jong Gierveld, 2004; Haskey, 2005; Lewis
and Haskey, 2006) as an emergent form of living arrangement. It is estimated
that there are some two million men and women in Great Britain who report
having a partner who lives in another household (Haskey, 2005). The BHPS
has collected information from sample members on non-co-residential partners,

but does not collect detailed information from or about these non-co-residential
partners. As people’s living arrangements and households become smaller and
more complex, their commitments and networks outside of the traditional ‘house-
hold’ tend to become greater, meaning that social science research needs to better
understand and reflect non-household-based definitions and sources of information
(ESRC, 2006).
There is a need for more finely grained qualitative research into the processes
underlying cohabiting unions, including their formation and dissolution. The vast
majority of research on cohabitation is based in the U.S. and is quantitative (Lewis,
2001). Large-scale, representative, quantitative datasets such as the BHPS give us
some clues as to potential avenues for further investigation. However, they cannot
fully account for the rapidly changing role of cohabitation in contemporary society.
There is an emergent body of qualitative research into cohabitation, including its
processes and meaning (Manning and Smock, 2005; Sassler, 2004; Lindsay, 2000).
There is a need to understand better what trends in cohabitation in particular, and
living arrangements in general, actually signify (Oppenheimer, 2003). Cohabitation
has emerged relatively recently and rapidly as a normative behaviour in many set-
tings, and is therefore in a situation of flux and change (Seltzer, 2000). The reasons
underlying decisions (whether articulated explicitly or otherwise) to cohabit may,
therefore, also be subject to rapid change, making cohabitation very much a moving
target to study.
Acknowledgment This research was based on the British Households Panel Survey Consolidated
Marital, Cohabitation and Fertility Histories, 1991–2006, produced by the Institute for Social and
Economic Research at the University of Essex. Access to the data was via the Economic and Social
Data Service, SN: 5629.
References
Alders, M.P.C. and Manting, D. (2001) Household scenarios for the European Union, 1995–202,
Genus, LVII (2): 17–48.
Allan, G. and Crow, G. (2001) Families, Households and Society, Palgrave, New York.
Axinn, W. and Thornton, A. (1993) Mothers, children, and cohabitation: the intergenerational
effects of attitudes and behaviour, American Sociological Review, 58 (2): 233–246.
Axinn, W.G. and Thornton, A. (2000) The transformation of the meaning of marriage. The ties
that bind: perspectives on marriage and cohabitation. Waite, L. (ed.), New York, Aldine de
Gruyter, pp. 147–165.
Bachrach, C., Hindin, M.J. and Thomson, E. (2000) The changing shape of ties that bind: an
overview and synthesis. The ties that bind: perspectives on marriage and cohabitation. Waite,
L., New York, Aldine de Gruyter, pp. 3–18.
Barlow, A., Duncan, S., James, G. and Park, A. (2001) Just a piece of paper? Marriage and cohab-
itation in Britain, in Park, A., Curtice, J., Thomson, K., Jarvis L. and Bromley, C. (eds.) British
Social Attitudes: The 18th Report – Public Policy and Social Ties, Sage, London.
Barlow, A. (2004) Regulation of cohabitation, changing family policies and social attitudes: a
discussion of Britain within Europe, Law and Policy 26(1): 57–86.
Bawin-Legros, B. and Gauthier, A. (2001) Regulation of intimacy and love semantics in couples
living apart together, International Review of Sociology 11(1): 39–46.
122 E. Coast
Beck-Gernsheim, E. (2002) Reinventing the Family: In Search of New Lifestyles, Polity Press,
Cambridge, pp. 170.
Bernard, J. and Bernard, J.S. (1982) The Future of Marriage, Yale University Press, New Haven.
Bernhardt, E. (2004) Cohabitation or Marriage? Preferred Living Arrangements in Sweden,
www.oif.ac.at/sdf/sdf04-04-bernhardt.pdf Accessed 30/06/08
Berrington, A. (2004) Perpetual postponers? Women’s, men’s and couple’s fertility intentions and
subsequent fertility behaviour, Population Trends, 112: 9–19.
Berrington, A. and Diamond, I. (2000) Marriage or cohabitation: a competing risks analysis of
first-partnership formation among the 1958 British birth cohort, Journal of the Royal Statistical
Society, 163(2): 127–152.
Berthoud, R. (2000) Introduction: the dynamics of social change. in Berthoud, R. and Gershuny, J.
(eds.) Seven years in the lives of Britain’s families: evidence on the dynamics of social change
from the British Household Panel Survey, Policy Press, Bristol, pp. 1–20.
Borell, K. and Karlsson, S.G. (2003) Reconceptualising intimacy and ageing: living apart together,
in Arber, S. Davidson, K. and Ginn, J. (eds.) Gender and Ageing: Changing Roles and Rela-
tionships, Open University Press, Maidenhead, pp. 47–62.
Bouchard, G. (2006) Cohabitation versus marriage: the role of dyadic adjustment in relationship
dissolution, Journal of Divorce and Remarriage, 46(1–2): 107–117.
Brown, S. (2000) Union transitions among cohabitors: the significance of relationship assessments
and expectations, Journal of Marriage and the Family, 62(3): 833–846.
Brown, S.L. and Booth, A. (1996) Cohabitation versus marriage: a comparison of relationship
quality, Journal of Marriage and the Family, 58(3): 668–678.
Bumpass, L.L. and Sweet, J.A. (1989) National estimates of cohabitation, Demography, 26:
615–25.
Bumpass, L.L., Raley, R.K. and Sweet, J.A. (1995) The changing character of stepfamilies: impli-
cations of cohabitation and non-marital childbearing, Demography, 32(3): 425–436.
Bumpass, L.L., Sweet, J.A. and Cherlin, A. (1991) The role of cohabitation in declining rates of
marriage, Journal of Marriage and the Family, 3(4): 913–927.
Burt, K.B. and Scott J. (2002) Parent and adolescent gender role attitudes in 1990s Great Britain,
Sex Roles, 46(7/8): 239–245.
Carlson, M., McLanahan, S. and P. England. (2004) Union formation in fragile families, Demog-
raphy, 41(2): 237–262.
Carter, W.Y. (1993) Attitudes towards pre-marital sex, non-marital childbearing, cohabitation, and
marriage among Blacks and Whites, National Survey of Families and Households Working
Paper 61, University of Wisconsin, Center for Demography and Ecology, pp. 62.
Casper, L.M. and Bianchi S.M. (2002) Continuity and Change in the American Family, Sage.
Cherlin, A. (2004) The deinstitutionalization of American marriage, Journal of Marriage and the
Family, 66(4): 848–861.
Cohan, C.L. and Kleinbaum S. (2002) Toward a greater understanding of the cohabitation effect:
premarital cohabitation and marital communication, Journal of Marriage and the Family, 64(1):
180–192.
Coontz, S. (2004) The world historical transformation of marriage, Journal of Marriage and the
Family, 66(4): 974–979.
de Jong Gierveld, J. (2004) Remarriage, unmarried cohabitation, living apart together: partner
relationships following bereavement or divorce, Journal of Marriage and the Family, 66(1):
236–243
DeMaris, A. (1984) A comparison of remarriages with first marriages on satisfaction in marriage
and its relationship to prior cohabitation, Family Relations, 33(3): 443–449.
de Vaus, D., Qu, L. and Weston, R. (2005) The disappearing link between premarital cohabitation
and subsequent marital stability, 1970–2001, Journal of Population Research, 22(2): 99–118.
Dey, I. and Wasoff, F. (2007) Protection, parity, or promotion: public attitudes to cohabitation and
the purposes of legal reform, Law and Policy, 29(2): 159–182
Ermisch, J. and Francesconi, M. (2000) Cohabitation in Great Britain: not for long, but here to
stay, Journal of the Royal Statistical Society, 163(2): 153–172.
ESRC (2006) Changing Household and Family Structures and Complex Living Arrangements,
ESRC Seminar Series: Mapping the Public Policy Lanscape, ESRC, Swindon.
Forste, R. (2002) Prelude to marriage or alternative to marriage? A social demographic look at
cohabitation in the U.S, Journal of Law and Family Studies, 4(1): 91–104.
Forste, R. and Tanfer, K. (1996) Sexual exclusivity among dating, cohabiting, and married women,
Journal of Marriage and the Family, 58(1): 33–47.
Fowers, B.J. (1991) His or her marriage: a multivariate study of gender and marital satisfaction,
Sex Roles, 24: 209–221.
Galland, O. (1997) Leaving home and family relations in France, Journal of Family Issues, 18(6):
645–670.
Gibson-Davis, C.M., Edin, K. and McLanahan, S. (2005) High hopes but even higher expectations:
the retreat from marriage among low-income couples, Journal of Marriage and the Family,
67(5): 1301–1312.
Giddens, A. (1992) The Transformation of Intimacy, Polity Press, Cambridge.
Gillis, J.R. (2004) Marriages of the mind, Journal of Marriage and the Family, 66(4): 988–991.
Haskey, J. (2001) Cohabitation in Great Britain: past, present and future trends – and attitudes,
Population Trends, 103, 4–25.
Haskey, J. (2005) Living arrangements in contemporary Britain: having a partner who usually lives
elsewhere and living apart together (LAT), Population Trends, 122: 35–45
Heuveline, P. and Timberlake, J.M. (2004) The role of cohabitation in family formation: the United
States in comparative perspective, Journal of Marriage and the Family, 66(5): 1214–1230.
Karlsson, S.G. and Borell, K. (2002) Intimacy and autonomy, gender and ageing: living apart
together, Ageing International, 27(4): 11–26.
Kenney, C.T. and McLanahan, S.S. (2006) Why are cohabiting relationships more violent than
marriages? Demography, 43(1): 127–140.
Kiernan, K. (2001) The rise of cohabitation and childbearing outside marriage in western Europe,
International Journal of Law, Policy and the Family, 15(1): 1–21.
Kiernan, K. (2004) Unmarried cohabitation and parenthood in Britain and Europe, Law and Policy,
26(1): 33–56.
King, V. and Scott, M.E. (2005) A comparison of cohabiting relationships among older and
younger adults, Journal of Marriage and the Family, 67: 271–285.
Knab, J. (2005) Cohabitation: Sharpening a Fuzzy Concept, Center for Research on Child Well-
being, Princeton University, Princeton.
Kravdal, O. (1999) Does marriage require a stronger economic underpinning than informal cohab-
itation? Population Studies, 53(1): 63–80.
Kwak, A. (1996) Non-marital cohabitation in law and public opinion in Poland, International
Journal of Law, Policy and the Family, 10(1): 17–26.
Lehrer, E.L. (2004) The role of religion in union formation: an economic perspective, Population
Research and Policy Review, 23(2): 161–185.
Lesthaeghe, R. (ed.) (2002) Meaning and Choice: Value Orientations and Life Course Decisions,
NIDI-CBGS publications No. 37, The Hague and Brussels.
Lesthaeghe, R. and Surkyn, J. (2004a) When history moves on: the foundations and diffusion of a
second demographic transition, Australian Population Association, 12th Biennial Conference,
15–17 September, Canberra.
Lesthaeghe, R. and Surkyn, J. (2004b) Value orientations and the second demographic transition
(SDT) in Northern, Western and Southern Europe: an update, Demographic Research, Special
Collection, 3(3): 45–86.
Levin, I. (2004) Living apart together: a new family form, Current Sociology, 52(2): 223–240.
Levin, I. and Trost, J. (1999) Living apart together, Community, Work and Family, 2(3): 279–294.
Lewis, J. (2001) Debates and issues regarding marriage and cohabitation in the British and Amer-
ican literature, International Journal of Law, Policy, and the Family, 15(1): 159–184.
Lewis, J. and Haskey, J. (2006) Living-apart-together in Britain: context and meaning, Interna-
tional Journal of Law in Context, 2(1): 37–48.
124 E. Coast
Liefbroer, A.C. and Dourleijn, E. (2006) Unmarried cohabitation and union stability: testing the
role of diffusion using data from 16 European Countries, Demography, 43(2): 203–221.
Lindsay, J.M. (2000) An ambiguous commitment: moving into a cohabiting relationship, Journal
of Family Studies, 6: 120–134.
Manning, W.D., Longmore, M.A. and Giordano, P.C. (2007) The changing institution of marriage:
adolescents’ expectations to cohabit and to marry, Journal of Marriage and the Family, 69(3):
559–575.
Manning, W.D. and Smock, P.J. (1995) Why marry? Race and the transition to marriage among
cohabitors, Demography, 32(4): 509–520.
Manning, W.D. and Smock P.J. (2002) First comes cohabitation and then comes marriage?
A research note, Journal of Family Issues, 23(8): 1065–1087.
Manning, W.D. and Smock, P.J. (2005) Measuring and modeling cohabitation: new perspectives
from qualitative data, Journal of Marriage and Family, 67(4): 989–1002.
Martin, C. and Thery, I. (2001) The Pacs and marriage and cohabitation in France, International
McGinnis, S.L. (2003) Cohabiting, dating and perceived costs of marriage: a model of marriage
entry, Journal of Marriage and the Family, 65(1): 105–116.
Milan, A. and Peters, A. (2003) Couples living apart, Canadian Social Trends, 69: 2–6.
Morgan, P.M. (2000) Marriage-lite: The Rise of Cohabitation and its Consequences, Institute for
the Study of Civil Society, London.
Mulder, C.H. and Manting, D. (1994) Strategies of nest-leavers: ‘settling down’ versus flexibility,
European Sociological Review, 10(2): 155–172.
Murphy, M. (2000a) Cohabitation in Britain, Journal of the Royal Statistical Society, 163(2):
123–126.
Murphy, M. (2000b) The evolution of cohabitation in Britain, 1960–95, Population Studies, 54(1):
43–56.
Mynarska, M. and Bernardi, L. (2007) Meanings and attitudes attached to cohabitation in Poland:
qualitative analyses of the slow diffusion of cohabitation among the young generation, Demo-
graphic Research, 16(17): 519–554.
Nock, S.L. (1998) Marriage in Men’s Lives, Oxford University Press, Oxford.
Office for National Statistics (ONS) (2007) Focus on Families: 2007 Edition,
[www.statistics.gov.uk/focuson/families].
Oppenheimer, V.K. (2003) Cohabiting and marriage during young men’s career-development pro-
cess, Demography, 40(1): 127–149.
Oropesa, R.S. (1996) Normative beliefs about marriage and cohabitation: a comparison of non-
Latino whites, Mexican Americans, and Puerto Ricans, Journal of Marriage and the Family,
58: 49–62.
Otnes, C. and Pleck, E.H. (2003) Cinderella Dreams: The Allure of the Lavish Wedding, University
of California Press, CA.
Prinz, C. (1995) Cohabiting, Married or Single: Portraying, Analyzing and Modelling New Living
Arrangements in the Changing Societies of Europe, Avebury, Aldershot.
Reynolds, J. and Mansfield, P. (1999) The effect of changing attitudes to marriage on its stability,
in Simons, J. (ed.) High Divorce Rates: The State of the Evidence on Reasons and Remedies,
Lord Chancellor’s Department Research Series, 2.99, London.
Rindfuss, R.R. and VendenHeuvel, A. (1990) Cohabitation: a precursor to marriage or an alterna-
tive to being single, Population and Development Review, 16(4):703–26
Sassler, S. (2004) The process of entering into cohabiting unions, Journal of Marriage and the
Family, 66(2): 491–505.
Schuman, H. and S. Presser, (1996) Questions and Answers in Attitude Surveys: Experiments on
Question Form, Wording, and Context, Sage, Thousand Oaks.
Scott, J., Alwin, D.F. and Braun, M. (1996) Generational changes in gender-role attitudes: Britain
in a cross-national perspective, Sociology, 30: 471–492.
Seltzer, J.A. (2000) Families formed outside of marriage, Journal of Marriage and the Family,
62(4): 1247–1268.
Seltzer, J.A. (2004) Cohabitation in the United States and Britain: demography, kinship, and the
future, Journal of Marriage and the Family, 66(4): 921–928.
Smart, C. and Stevens, P. (2000) Cohabitation Breakdown, London Policy Studies Centre,
London.
Stanley, S.M., Rhoades, G.K. and Markman, H.J. (2006) Sliding versus deciding: inertia and the
premarital cohabitation effect, Family Relations, 55(4): 499–509.
Sweet, J.A. and Bumpass, L.L. (1992) Young adults’ views of marriage, cohabitation and family, in
South, S.J. and Tolany, S.E. (eds.) The Changing American Family, Westview Press, Boulder,
pp. 143–170.
Teitler, J.O. and Reichman, N.E. (2001) Cohabitation: an elusive concept, Social Indicators Survey
Center Working Paper 01-4, Columbia University.
Théry, I. (1993) Le Démariage, Justice et Vie Privée, Odile Jacob, Paris.
Thomson, E. and Collela, U. (1992) Cohabitation and marital stability: quality of commitment,
Journal of Marriage and the Family, 54(2): 259–267.
Thornton, A. (1989) Changing attitudes toward family issues in the United States, Journal of Mar-
riage and the Family, 51(4): 873–893.
Thornton, A. (1995) Attitudes, Values and Norms related to Non-marital Fertility. Report to
Congress on out-of-wedlock Childbearing, DHSS pub no {PHS} 95-1257, US Gov Printing
Office, pp. 201–215.
Thornton, A., Axinn, W.G. and Hill, D.H. (1992) Reciprocal effects of religiosity, cohabitation,
and marriage, American Journal of Sociology, 93(3): 628–651.
Thornton, A. and Young-DeMarco, L. (2001) Four decades of trends in attitudes toward family
issues in the United States: the 1960s through the 1990s, Journal of Marriage and the Family,
63(4): 1009–1037.
Tourangeau, R., Rips, L.J. and Rasinski, K. (2000) The Psychology of Survey Response, Cambridge
University Press, Cambridge.
Trost, J. (1978) A renewed social institution: non-marital cohabitation, Acta Sociologica, 21(4):
303–315.
University of Essex (2006) British Household Panel Survey: Waves 1–15. Institute for Social and
Economic Research, SN5151.
Van de Kaa, D.J. (1987) Europe’s second demographic transition, Population Bulletin, 42(1):
1–59.
Waite, L. and Gallagher, M. (2000) The Case for Marriage: Why Married People Are Happier,
Healthier, and Better Off Financially, Broadway, New York.
Chapter 7
Living Arrangements, Health and Well-Being
Harriet Young and Emily Grundy
Introduction
Populations throughout Europe and the developed world are ageing, with growing
proportions over age 60, associated with increases in life expectancy and falling
birth rates. Between 1960 and 2004, the proportion of people aged 65 and over
in the UK increased by a third from 11.7 to 15.6%, and the proportion of the
population aged 85 and over nearly tripled (Soule et al., 2005). In the same time
period, there have been major changes in living arrangement patterns of older
people throughout Europe and the developed world. Older people are much less
likely to live with relatives in multi-generational households than previously and
are much more likely to live alone (Elman and Uhlenberg, 1995; Glaser et al., 2004;
Iacovou, 2000). For example, in Italy the proportion of women aged 65 and over
living alone was 22% in 1971, and had risen to 36% by 2000 (Tomassini et al.,
2004). In the last decade, there has been a stalling of the trend towards increased
solitary living among older people in Europe, but this has been driven by demo-
graphic changes, notably increases in the proportion of older people who are mar-
ried, rather than by changes in the residence patterns of the unmarried (Grundy,
1996; Tomassini et al., 2004). The overall pattern, however, is of greater residential
independence among older people with larger proportions living alone or just with a
spouse.
These changes in the living arrangements of older people reflect, it has been
hypothesised, a number of factors, including attitudinal shifts towards a greater
desire for privacy and individualisation, especially in northern Europe (De Jong
Gierveld et al., 2001; Pampel, 1992). Economic changes which have led to increas-
ing financial independence coupled with improvements in health status have, it
is argued, allowed more older people to realise their desires for independence
E. Grundy (B)
Centre for Population Studies, London School of Hygiene and Tropical Medicine,
University of London, UK
e-mail: Emily.Grundy@lshtm.ac.uk

128 H. Young and E. Grundy
(Glaser et al.,1997; McGarry and Schoeni, 2000; Michael et al., 1980). Other factors
include demographic changes such as increased life expectancy and changing age
at completion of childbearing (Grundy, 2006).
Despite the common direction of this shift throughout Europe, there are still
significant differences between European countries in living arrangement patterns.
In southern Europe, older people are still more likely to live in multi-generational
households and less likely to live alone than in northern Europe. In Spain, the pro-
portion of older people living with relatives decreased from 58% in 1970 to 23%
in 1993, but in Sweden in the mid 1990s, only 2% of older people lived with chil-
dren (Sundstrom and Tortosa, 1999). These differences may be associated with cul-
tural and attitudinal variation across Europe, and differences in welfare regimes and
family-related policies (Daatland and Herlofson, 2003; Millar and Warman, 1996;
Reher, 1998).
Much of the debate about these trends has centred on the public policy implica-
tions of possible reductions in the availability of family support for older people at
a time when the numbers and proportions of older Europeans are rapidly increas-
ing (European Commission, 2006). Equally important are the implications for the
health and well-being of the older population, as recognised in the Madrid Plan of
Action adopted at the 2002 Second World Assembly on Ageing. This called for
more research on the advantages and disadvantages of different living arrangements
for older people (United Nations, 2002). In England, a study of risk factors in social
exclusion found that those living alone are likely to experience exclusion on mul-
tiple dimensions (ODPM, 2006) which may well have implications for health and
well-being. Indeed, older people living alone in England have been identified as a
potential ‘at risk’ group for mental health problems (Department of Health, 2001).
In this chapter, we investigate the association between living arrangements and
health and well-being in England and Wales and make comparisons with data on
other European countries. We first describe living arrangement patterns in England
and Wales and in Europe. Second, we examine in detail the associations between
living arrangements and health and emotional well-being in England, and health
and mortality in England and Wales. Using longitudinal data, we additionally
examine the possible influence of health selection on our results. We also make
comparisons with analyses of European data on associations between living arrange-
ments and health and well-being, to observe the moderating effect that different
cultural and political environments may have on these associations.
Previous Research
Research generally suggests that older people who live with a spouse have lower lev-
els of mortality, better physical health (Kendig et al., 2007; Koskinen et al., 2007;
Murphy et al., 2007) and higher levels of emotional well-being than those not living
with a spouse (Kendig et al., 2007; Kohler et al., 2005; Mindel and Wright, 1982).
Living with a spouse may provide emotional intimacy, economic benefits, social
7 Living Arrangements, Health and Well-Being 129
control of behaviour, and more opportunities for social integration (Hughes and
Waite, 2002; Murphy et al., 2007; Pillemar et al., 2000), all of which are likely
to influence health and well-being.
For unmarried older people living independently (rather than in communal estab-
lishments), choices revolve around living with relatives and friends or living alone.
A number of studies from England, Spain, Canada and the USA report that older
people living alone are more likely than others to report poor health status, low lev-
els of psychological health and quality of life than those living with others (Burnette
and Mui, 1996; Garcia et al., 2005; Hall and Havens, 2001; Kharicha et al., 2007).
One study from Denmark found that women aged 75 and over who lived alone had
higher mortality risks than those living with others (Lund et al., 2000). There are
a number of reasons why living alone might be associated with poorer health and
well-being than living with others. Older people living alone are often portrayed
as a vulnerable group who need support, and who are potentially at risk of poorer
mental and physical health (Barnes et al., 2006; Kharicha et al., 2007). One illustra-
tion of such vulnerability was the elevated risk of death among older people living
alone in the heat waves in Paris in 2003 and in Chicago in 1995 (Crumley, 2003;
Klinenberg, 2002). There are also a number of potential benefits of living with oth-
ers rather than alone for health and well-being. Older people spend larger amounts
of time at home than others and the household is an important context in which both
financial and social support, companionship and care are provided and exchanged
(De Jong Gierveld et al., 2001; Grundy, 1999; Hahn, 1993).
Opposing these results, however, is a substantial body of research from the UK
and the USA which indicates that older people living alone experience better health,
lower levels of loneliness and depression and higher quality of life compared with
the unmarried living with others (Gustavson and Lee, 2004; Iliffe et al., 1992;
Michael et al., 2001; Netuveli et al., 2006; Sarwari et al., 1998; Wenger, 1984).
Lower levels of mortality among those living alone compared with those living with
others have also been reported in Japan and Italy (Murata et al., 2005; Pizzetti et al.,
2005; Walter-Ginzburg et al., 2002). These findings may in part be explained by the
fact that the majority of older people living alone are not isolated and have exten-
sive social contacts and support from family and others outside the household, and
they may prefer to be independent of support from children for as long as possible
(Daatland and Herlofson, 2001; Silverstein and Bengston, 1994; Ulbrich and
Warheit, 1989). Living with others rather than alone may also contribute to poorer
health status. The return home of children needing support may be stressful and
therefore damaging to health (Grundy, 2000). Higher levels of loneliness among
unmarried older people living with family have also been reported, due to an
increased sense of obligation, and a loss of privacy and self-determination (De Jong
Gierveld et al., 2001;Wenger, 1984).
Associations between living alone and better health may also be an artefact of
health selection. A health selection hypothesis would suggest that those living with
others do so because of poor health or disability, and that those still living alone
are able to do so because they are in good health and can continue to support them-
selves, as has been demonstrated by previous research (Brown et al., 2002; Grundy,
1993; Mui and Burnette, 1994). Health selection is one of the reasons hampering
an understanding of the effects of living arrangements on health and this may at
least in part explain the mixed findings of the research mentioned above. How-
ever, results from different longitudinal studies in the USA which control for initial
health status still produce conflicting results (Michael et al., 2001; Mor et al., 1989;
Sarwari et al., 1998).
Another important consideration when investigating the association between liv-
ing arrangements and well-being is that the health implications of a particular liv-
ing arrangement are likely to be mediated by other factors including cultural and
socio-economic factors, the availability of extra-household resources and individual
factors such as domestic skills (Grundy, 2001). There is evidence that cultural varia-
tion influences the association between living arrangements and well-being. Studies
of older people in Spain, for example, have shown that older women living with
their children report higher levels of satisfaction than other women (Garcia et al.,
2005; Zunzunegui et al., 2001). In contrast, a study conducted in Wales found that
older people living with relatives were the most likely to report loneliness and poor
morale (Wenger, 1984). There is also evidence that the possible consequences of
living alone depend on the extent of extra-household social support and interaction.
Fratiglioni found that, in Sweden, older people living alone who had no satisfying
contacts outside the household had a higher risk of cognitive decline than those
with such contacts (although risks of cognitive decline were higher among those
living alone with extra household interaction than among those living with others)
(Fratiglioni et al., 2000). Wealth may also influence the association between living
arrangements and health status (Ulbrich and Warheit, 1989), and financial support
and transfers from co-residents may be important for health outcomes for those on
low incomes (Grundy, 2001).
Data and Methods
Data Sources and Study Populations

We used three datasets for our analyses, the English Longitudinal Study of Age-
ing (ELSA) with data on England only; the Office for National Statistics (ONS)
Longitudinal Study (LS) with data on England and Wales; and the European Social
Survey (ESS) with data on 19 European countries. For all datasets, we defined our
study population as those aged 60 and over who were living in the community. The
LS included data on those living in communal establishments (defined as estab-
lishments providing managed accommodation with full- or part-time supervision,
for example hospitals, care homes, prisons, and sheltered accommodation where at
least half of residents do not possess their own cooking facilities (Office for National
Statistics, 2004)) but the other two datasets did not, and so we excluded this popula-
tion from our LS analyses. Note that LS analysis including those living in communal
establishments in 2001 did not alter the results presented.
The ELSA was initiated in 2002, and the original sample was selected from
respondents of the Health Survey for England (HSE) in 1998, 1999 and 2001, each
of which had response rates of between 67 and 70% (Marmot et al., 2002). To date,
three waves of ELSA data have been collected and in our analysis we use data from
the first two waves, which had response rates of 67 and 82% respectively. We used
a cross-sectional sample of 7,226 individuals present at the first wave, and another
sample of those who were still in the study at wave two, consisting of 5,034 sample
members. Of the wave one sample not present at wave two, 54% refused to answer,
26% had died since the first wave, and the remainder were ill and unable to answer,
had entered a communal establishment or were lost to follow up (Banks et al., 2006;
Marmot et al., 2002).
The ONS LS is a record linkage study of 1% of the population of England
and Wales, initially based on those enumerated at the 1971 Census (approximately
500,000 people) (Hattersley and Creeser, 1995). Sample members were selected on
the basis of birthday and the sample has been maintained by recruitment of new
births and immigrants born on LS birthdays. Record linkage has been used to add to
the dataset information from subsequent censuses (1981, 1991, 2001) and from vital
registration including death of a spouse and death. Census information on all house-
hold members at each census point is included in the dataset including information
on the relationship to the LS member. We used a sample of 78,751 respondents aged
60 and over at the 2001 Census and also present at the previous three census points.
Sample members had to be present at the 1971 and 1981 Census points so that we
could establish their socio-economic status when they were of employment age, and
had to be present at the 1991 Census for longitudinal analyses of changes in living
arrangement.
The ESS is a biennial social survey, and to date three rounds have been car-
ried out in 2002, 2004 and 2006, including 22, 26 and 25 countries respectively.
Samples in each country were selected randomly to ensure representativeness of
the population. A strength of the ESS is that there are clear and detailed cen-
tral survey specifications which all country studies adhere to, and close collabo-
ration on protocols to ensure correct translations to multiple languages (Jowell,
2003). We used the first two rounds of data in our analysis, and pooled them,
selecting 19 countries for analysis. We excluded certain countries (Italy, France,
Luxembourg, Switzerland) with response rates below 50% in either round; Ire-
land because it was anomalous in terms of living arrangements and health status
with respect to our country groupings discussed below and Iceland because of dif-
ferences in welfare regime to other Scandinavian countries (Olafsson, 2003). The
resulting sample comprised 17,770 people aged 60 and over in 2002 or 2004. We
were not able to carry out detailed country-specific analysis due to small individual
country sample sizes. Instead we made four regional groupings: northern, west-
ern, southern and eastern Europe (Table 7.1), on the basis of established typologies
of predominant family cultures, welfare regimes, and geographic location (Kohli
and Albertini, 2006; Millar and Warman, 1996). To check the appropriateness of
these groupings, we examined the distributions of living arrangements, health, hap-
piness and satisfaction with life by country within region, to ensure reasonable
Table 7.1 Regional groupings for European countries used in analysis of the ESS
North West South East
Denmark Austria Greece Czech Republic

Finland Belgium Portugal Estonia
Norway Germany Spain Hungary
Sweden Netherlands Poland
UK Slovakia
Slovenia
Ukraine
n = 3,621 n = 5,867 n = 3,857 n = 4,786
homogeneity within regional groupings. This satisfied us that these regional group-
ings were appropriate.
Variables
Outcome variables included depression and loneliness (ELSA), happiness (ESS),

self-rated health (all three datasets), presence of a limiting long-term illness (LLTI)
and an indicator of death between the 2001 Census and 2004 (LS). We describe these
in more detail in the relevant results sections. In analyses of all three datasets, we
used a living arrangement variable with three categories including those living with
a spouse and, for the unmarried, those living alone and those living with others. The
unmarried group included those who were never-married, divorced, separated and
widowed. For longitudinal analysis of the LS, we developed a variable measuring
change in living arrangement between 1991 and 2001, focusing on those not living
with a spouse in 1991. This variable had five categories including two categories of
no change (living alone at both time points; living with others at both time points)
and three categories of change (changing to live with a spouse; changing to live
alone; and changing to live with others). Other co-variates used in analysis included
age and various indicators of socio-economic status in all three datasets; presence
of a LLTI for analyses of emotional well-being using the ESS and ELSA and health
selection using the LS; and region for analyses of the whole European sample using
the ESS.
Analysis Methods
In the first part of our analysis, we examined living arrangement patterns in England
and Wales and the other European countries included in the ESS sample. We then
analysed the associations between living arrangements and health and well-being,
first examining associations descriptively, and then using multivariate methods to
model associations controlling for other co-variates. We used logistic regression to

model binary outcomes including presence of loneliness, depression, poor self-rated
health and, for analysis of the LS, death between the 2001 Census and the end of
2004.
The outcome measure for logistic regression models is the odds ratio (OR).
This is a measure of association between two factors and compares whether the
probability of a certain event is similar for two groups. In Table 7.2, for exam-
ple, the odds ratio for the model with the outcome of depression for men com-
paring those living with a spouse to living alone is 0.41, which implies that
men living with a spouse are 0.41 times as likely (i.e. 59% less likely) to be
depressed than men living alone. We also used ordinal logistic regression to model
happiness, an outcome with an ordinal scale, in ESS analysis. The odds ratio
from ordinal logistic regression represents the odds of being in a higher cate-
gory than a lower category (e.g. a score of 10 rather than 9; 9 rather than 8
et cetera on the happiness scale). A higher odds ratio therefore indicates higher
levels of happiness.
Results
Living Arrangements in England and Europe
Figures 7.1 and 7.2 show the living arrangement distributions of the samples
for each dataset separately by gender. The results from ELSA and the LS are,
unsurprisingly, almost identical and show that in England and Wales, approximately
With others Alone With spouse

100
Percentage of men aged 60+
80
60
40
Fig. 7.1 Distribution of men
aged 60+ by living 20
arrangements in England,
England and Wales and
regions of Europe 0
England & Northern Western Southern Eastern
(Source: ELSA, 2002; LS, England
Wales Europe Europe Europe Europe
2001; ESS, 2002 and 2004;
authors’ analysis.) ELSA ONS LS ESS
Fig. 7.2 Distribution of With others Alone With spouse

women aged 60+ by living 100
Percentage of women aged 60+

arrangements in England,
England and Wales and
regions of Europe 80
(Source: ELSA, 2002; LS,
2001; ESS, 2002 and 2004;
authors’ analysis.) 60
40
20
0
England & Northern Western Southern Eastern
England
Wales Europe Europe Europe Europe
ELSA ONS LS ESS
three quarters of older men and half of older women lived with a spouse. More
detailed analyses (not shown) indicated that a vast majority lived only with a spouse
and that those living with a spouse and others were most likely to live with a spouse
and children. Of those not living with a spouse in our ELSA and LS samples, the
majority of both men and women lived alone. In the LS sample, 41% of women and
18% of men aged 60 and over lived alone. Among those aged 75 and over, propen-
sity to live alone was even higher, at 60% for women and 29% for men, largely
reflecting the higher likelihood of being widowed among the oldest old. The gender
difference reflects the higher life expectancy of women than men, combined with
the fact that women tend to marry men older than themselves. The ESS data demon-
strate variation in living arrangements by region of Europe, consistent with a large
body of other research (Iacovou, 2000; Reher, 1998; Tomassini et al., 2004). The
pattern for England and Wales mirrors that of the western European region (which
includes the UK). In the northern region, older unmarried people were the most
likely to live alone and in eastern and southern Europe, they were much more likely
to live with others. These differences were still evident in multivariate analysis (not
shown) in which we controlled for age and socio-economic status.
Living Arrangements, Depression and Loneliness in England
In the following sections, we consider the associations between living arrangements

and health and well-being in England, and England and Wales. We focus first on
emotional well-being and second on self-rated health and mortality, and make com-
parisons with analysis of European data.
Using ELSA, we examined the association between living arrangements and

two outcome variables, depression and loneliness. In ELSA, depression was mea-
sured using the Centre for Epidemiological Studies eight-item scale of depressive
symptoms and psychological distress (CES-D) (Radloff, 1977; Steffick, 2000). The
CES-D scale measures a continuum of depression rather the presence or absence of
a psychiatric disorder (Steffick, 2000), but for the purposes of this chapter, we used
a cut-off point of three or more symptoms to indicate a high likelihood of depres-
sion, a convention used in other research (Comstock and Helsing, 1976, and refer
to this group as having depression. In our sample, 26% had at least three symptoms
of depression and this was more likely for women than men (30 and 21% respec-
tively) and for those aged 75 and over than 60–74 year olds (for women, 36 and 26%
respectively). Data on loneliness was collected at wave two only, using a four item
scale (Hughes et al., 2004), based on the 20-item Revised UCLA loneliness scale
(Russell, 1996). We collated this into a summary score of zero to nine based on
addition of scores (0–2) from each item. We developed a binary measure and those
with scores of 5 and above were considered to be lonely. Women were more likely
to report loneliness than men (17 compared with 13%), and for men only, loneliness
was more common among those aged 75 and over compared with those aged 60–74
(16 compared with 11%).
We calculated prevalences of depression at wave one and loneliness at wave two
by living arrangement at wave one. We then carried out multivariate analysis using
logistic regression to examine associations between living arrangements and the
outcome variables controlling for age, socio-economic status and for presence of
illness, all of which have been shown to be associated with emotional well-being
(Braam et al., 2005; Montgomery et al., 2007; Wilson et al., 1999). Analysis of
loneliness as an outcome excluded those who were lost to follow up by wave two. To
check whether this was likely to bias results, we compared our depression analysis
which used wave one data with analysis of depression score at wave two as an
outcome, and results were similar, and so we do not expect a major bias in loneliness
analysis.
Men and women living alone had the highest prevalence of depression, while
those living with a spouse had the lowest (Fig. 7.3), particularly among men. A
similar pattern was evident for loneliness, and among men 22% of those living
alone had high loneliness scores compared with only 10% of those living with
a spouse. When controlled for age, presence of an illness and socio-economic
status, women living with a spouse had a significantly lower likelihood of being
lonely and depressed than those living alone. Those living with others were also
less likely to be depressed and lonely than those living alone, although only
significantly so for women (of borderline significance for depression) (Table 7.2).
Other factors associated with depression and loneliness in our models were
presence of a LLTI, lower levels of wealth, and for depression only, lower levels of
education.
In the next section, we examine the associations between living arrangements and
emotional well-being in Europe, to ascertain how other areas of Europe compare
with England and Wales in this regard.
Fig. 7.3 Proportion of the 50
Percentage of population aged 60+

With spouse Alone With others
population aged 60 and over
with high depression scores
by living arrangement and 40
gender, England
(Source: ELSA, 2002; 30
authors’ analysis.)
20
10
0
Men Women
Living Arrangements and Happiness in Europe
In the ESS, a measure of happiness was derived from responses to the questions
‘taking all things together, how happy would you say you are?’ Respondents rated
their answer on a scale of zero (extremely unhappy) to ten (extremely happy). For
descriptive analysis, we derived a binary measure, and those with scores below
6 were considered to be unhappy. For multivariate analysis using ordinal logistic
regression, we derived a six category variable (0 to 4, 5, 6, 7, 8 and 9 to 10), as very
few respondents gave answers at the top or bottom of the range.
We observed striking differences in happiness between the 19 European countries
included in our sample. The northern region had the highest average happiness
scores, followed by the west, south and finally the eastern region, and this gradient
can be observed in Fig. 7.4. At the country level, average happiness scores were the
highest in Denmark and lowest in the Ukraine, and Great Britain had the fifth high-
est average score. These country differences in mental health have also been well-
documented elsewhere (Castro-Costa et al., 2007; Lehtinen et al., 2005), although
cross-cultural variability in response to questions in the CES-D scale have been
reported, which may also have affected our findings (Iwata and Buka, 2002).
In all regions of Europe, those living with a spouse had the lowest prevalence of
unhappiness (Fig. 7.4). In the southern region, those living alone had a higher preva-
lence of unhappiness than those living with others among both men and women, but
in other regions, findings were mixed. It is worth noting here that regional differ-
ences were so distinct that in the eastern region, even men living with a spouse
reported a higher prevalence of unhappiness than those in all other living arrange-
ment groups in other regions.
We next used ordinal logistic regression to ascertain whether differences were
significant after control for age and socio-economic status. These analyses demon-
strated that in all regions, those living with a spouse were significantly happier than
those living alone (results not shown). Among unmarried women for all regions
Table 7.2 Results from logistic regression model of proportion of men and women with depression
in 2002 and loneliness in 2004 by living arrangements and other factors in 2002, England
Men Women
Odds ratio p-value 95% CI Odds ratio p-value 95% CI
Depressiona
Living Alone 1.00 1.00
arrangements With spouse 0.41 ∗∗∗ 0.33,0.51 0.76 ∗∗ 0.63,0.91
With others 0.77 0.47,1.25 0.75 0.56,1.01
Age 1.00 0.98,1.01 1.01 1.00,1.02
LLTI No 1.00 1.00
Yes 3.89 ∗∗∗ 3.20,4.74 2.98 ∗∗∗ 2.56,3.47
Net wealth Less wealth 1.19 ∗∗∗ 1.10,1.29 1.18 ∗∗∗ 1.10,1.26
Housing tenure Owner 1.00 1.00
occupier
Renter 1.18 0.93,1.49 1.18 0.97,1.42
Highest Higher 1.00 1.00
educational secondary
qualification or above
Lower 1.05 0.79,1.40 1.05 0.81,1.37
secondary
None 1.47 ∗∗ 1.15,1.87 1.58 ∗∗∗ 1.31,1.90
n = 3,093 n = 3,739
Lonelinessb
arrangements With spouse 0.44 ∗∗∗ 0.33,0.60 0.47 ∗∗∗ 0.37,0.59
With others 0.52 0.25,1.06 0.66 ∗ 0.45,0.98
Age 1.01 1.00,1.03 0.98 ∗ 0.97,1.00
LLTI No 1.00 1.00
Yes 1.85 ∗∗∗ 1.42,2.39 1.69 ∗∗∗ 1.38,2.08
Net wealth Less wealth 1.09 0.98,1.22 1.14 ∗∗ 1.04,1.24
occupier
Renter 1.09 0.77,1.53 0.82 0.63,1.07
Lower 1.24 0.88,1.75 1.04 0.76,1.42
secondary
None 1.16 0.85,1.58 1.14 0.90,1.44
n = 2,395 n = 2,997
a
Comparing those with three or more symptoms of depression to those with two or fewer symp-
toms. b Comparing those with a score of 5–9 to those with a score of less than 5. ∗ p<0.05
∗∗
p<0.01∗∗∗ p<0.001.
Source: ELSA, 2002; authors’ analysis.
combined, those living with others were significantly happier than those living
alone. However, analysis by region demonstrated a significant association in the
eastern, southern and western regions (supporting our findings from ELSA), but not
in the northern region.

100
80
60
40
20
0
North West South East North West South East
Men Women
Fig. 7.4 Proportion of older people with low happiness scores by European region and gender
(Source: ESS 2002, 2004; authors’ analysis.)
Among men, we found no significant associations between living arrangements

and happiness for the unmarried in any region or for all regions combined, in accord
with our findings from ELSA. The gender difference in our results is consistent
with the results of some studies which indicate that it is the presence of a partner
that is important for men’s mental health whereas for women, alternative bonds,
such as with children, may more often offer an alternative (De Jong Gierveld, 2003;
Josselson, 1996). However, this finding for men is also in contrast to a few single
country studies (Grundy, 1989; Zunzunegui et al., 2001), and it is likely that small
sample sizes, especially of those living with others, limited the statistical power
of our analyses. We next examined the association between living arrangements
and self-rated health and mortality. Self-rated health is a general indicator of health
status, encompassing both psycho-social and physical dimensions of health.
Living Arrangements, Self-rated Health and Mortality

in England and Wales
The questions on self-rated health were different in ELSA and the LS, as were the
response categories. For both datasets, we developed a binary self-rated health vari-
able, and aimed to include a quarter of the sample in the poorer self-rated health
category; the categories achieved represent a best-fit. In ELSA, respondents were
asked to rate their health as excellent, very good, good, fair or poor. These cate-
gories were aggregated into a binary variable with poor and fair versus excellent,
very good and good self-rated health, with 32% of the sample in the poorer health
category. In the LS, answers to the question on self-rated health at the 2001 Census
were categorised as good, fair or poor. These were aggregated into good and fair
versus poor health, and 21% of the sample had poor health. Our other LS outcome
variable was an indicator of death between the 2001 Census and the end of 2004,
and 13% of the sample died during this follow up period. For men and women in
both the LS and ELSA samples, the prevalence of poor self-rated health was lowest
among those living with a spouse, intermediate for those living alone and was high-
est for the unmarried living with others (Fig. 7.5). The prevalence of mortality was
lowest among those living with a spouse, but the difference between those living
alone and with others was marginal.
Logistic regression was used to examine this association controlled for age and
socio-economic status, both of which are known to be associated with health status
and with living arrangements (Table 7.3). Among the unmarried, analysis of the LS
indicated that those living with others were more likely to report poor health and had
higher risks of mortality than those living alone. Odds ratios from analysis of ELSA
also indicated that living with others was associated with poor health, but results
were not significant. This may reflect small sample sizes of those living with oth-
ers, which reduced the statistical power of the analysis. Rather surprisingly, women
living with a spouse were significantly more likely to rate their health as poor than
those living alone after control for age and socio-economic status in both the LS
and ELSA. Further investigation of this association in the LS by age demonstrated
that this association was only present for women aged 75 and over. This finding is
consistent with research from the UK and the USA, which also shows that older
old women living alone report better health than those living with a spouse. This
may reflect the fact that this group includes the never-married who have stronger
social bonds with others outside the household (Goldman et al., 1995; Grundy and
Sloggett, 2003), and that they do not have to provide support and care for a spouse,
50
40
30
20
10
0
Men Women Men Women Men Women
Self-rated health 2002 Self-rated health 2001 Mortality 2001-04

ELSA ONS LS
Fig. 7.5 Prevalence of poor self-rated health and mortality for the population aged 60+ by living
arrangement and gender
(Source: ELSA, 2002; ONS LS, 2001–2004; authors’ analysis.)
Table 7.3 Results from logistic regression model of proportion of men and women aged 60+
with poor self rated health 2001/2002 and mortality 2001–2004 by living arrangements and other
factors, England 2002, England and Wales 2001
Men Women
Odds Ratio p-value 95% CI Odds Ratio p-value 95% CI
ELSA: Poor self-rated health 2002a

arrangements With spouse 1.11 0.91,1.36 1.25 ∗ 1.05,1.49
With others 1.17 0.75,1.83 1.10 0.82,1.47
Age 1.01 ∗ 1.00,1.02 1.02 ∗∗∗ 1.01,1.03
Net wealth Less wealth 1.28 ∗∗∗ 1.20,1.37 1.34 ∗∗∗ 1.26,1.43
occupier
Renter 1.78 ∗∗∗ 1.46,2.18 1.37 ∗∗∗ 1.14,1.64
Lower 1.06 0.84,1.34 0.98 0.75,1.27
secondary
None 1.64 ∗∗∗ 1.35,1.99 1.49 ∗∗∗ 1.24,1.79
N = 3,176 N = 3,837
LS: Poor self-rated health 2001b
arrangements With spouse 0.98 0.91,1.04 1.16 ∗∗∗ 1.10,1.23
With others 1.30 ∗∗∗ 1.14,1.48 1.36 ∗∗∗ 1.25,1.47
Age 1.03 ∗∗∗ 1.03,1.03 1.05 ∗∗∗ 1.05,1.05
Region South & East 1.00 1.00

Central 1.30 ∗∗∗ 1.22,1.38 1.38 ∗∗∗ 1.31,1.46
North & 1.58 ∗∗∗ 1.48,1.68 1.51 ∗∗∗ 1.43,1.60
Wales
Tenure and car Lower score 1.24 ∗∗∗ 1.22,1.27 1.24 ∗∗∗ 1.22,1.26
access c
Social class 1971 1.18 ∗∗∗ 1.14,1.23

N = 35,689 N = 43,062
LS: Mortality 2001–2004
arrangements With spouse 0.85 ∗∗∗ 0.79,0.92 0.93 0.87,1.01
With others 1.11 0.96,1.30 1.24 ∗∗∗ 1.13,1.37
Age 1.12 ∗∗∗ 1.12,1.13 1.12 ∗∗∗ 1.12,1.12
Region South & East 1.00 1.00

Central 1.06 0.99,1.15 1.04 0.97,1.12
Men Women
Odds Ratio p-value 95% CI Odds Ratio p-value 95% CI
North & 1.07 0.99,1.16 1.12 ∗∗ 1.04,1.21

Wales
Tenure and Lower score 1.17 ∗∗∗ 1.14,1.20 1.14 ∗∗∗ 1.12,1.16
car access c
Social class 1.03 0.98,1.08

1971
N = 35,689 N = 43,062
a
Outcome variable self rated health categorised into very poor and poor versus fair, good
and very good health. b Outcome variable self rated health categorised into not good versus
fairly good and good health. c Summary score of tenure and car access in 1971 and 1981.
∗
p<0.05∗∗ p<0.01∗∗∗ p<0.001.
Source: ELSA, 2002; ONS LS, 2001–2004; authors’ analysis.
with possible negative implications for health and well-being. However, the never-
married only accounted for 13% of those living alone, and the majority were wid-
owed. In opposition to our finding for self-rated health, women living with a spouse
had a lower (of border line significance) odds of mortality compared with those
living alone. These differences by type of health indicator show that different mea-
sures are indicators of different processes. For example, self-rated health captures
socio-psychological dimensions of health (Grundy and Sloggett, 2003; Idler and
Benyamini, 1997). Our model results also showed that increasing age, having lower
levels of socio-economic status, and living outside the South and East of England
were associated with poorer self-rated health and higher mortality.
Our results consistently indicated that among the unmarried, those living with
others were more likely to have poor self-rated health and higher mortality than
those living alone. This finding may be a result of health selection and we next
examine the part that health selection may have played in our results using the LS.
To do this, we examined the association between living arrangement change from
1991 to 2001 health outcomes in 2001 for a subpopulation of our LS sample. We
focused on LS members in our sample who were unmarried and did not report a
limiting long term illness (LLTI) at the 1991 Census. We excluded those who were
living with a spouse in 1991 in order to exclude the health influence of widowhood
in the inter-censal period on our results (Lee et al., 2001; Wilcox et al., 2003). The
sample may still have included individuals who were unmarried in 1991 but then
were married and widowed before 2001, but this proportion would be small in the
age groups we consider. We excluded those with a LLTI in 1991 because we were
interested in health related changes in living arrangement. We made the assump-
tion that individuals who did not have an LLTI in 1991 and did in 2001 and who
changed to living with others were likely to have made a health-related change.
If those who changed to living with relatives rather than living alone or living
with others at both time points were more likely to have developed poor self-rated
health in 2001, we assumed that this would demonstrate a health selection effect on
our results.
We used logistic regression to model the association between living arrange-
ment change and all three health outcome variables, controlling for age and socio-
economic status (Table 7.4). This analysis confirms that those changing to live with
others were significantly more likely to have an LLTI and poor self-rated health
in 2001 than those who lived alone at both time points. They also had increased
odds of death 2001–2004, but this association was not significant. Among women,
those living with others in both 1991 and 2001 were also more likely to experience
poor health outcomes in 2001, but the magnitude of the odds ratios was marginally
Table 7.4 Results from logistic regression model of proportion of men and women who were
unmarried and who had no LLTI in 1991, who then had poor self rated health or an LLTI in
2001 or died 2001–2004 by living arrangement change 1991–2001, controlled for other factors,
a
England and Wales
Men Women
Living arrangement OR p- 95% CI N OR p- 95% CI N
change 1991–2001 value value
Self-rated health
No change Alone – alone 1.00 1,980 1.00 6,638
With others – with 1.15 0.90,1.46 615 1.39 ∗∗∗ 1.21,1.59 1,697
others
Change Alone – with others 2.29 ∗∗∗ 1.40,3.74 85 1.67 ∗∗∗ 1.27,2.19 312
With others – alone 1.03 0.80,1.33 637 0.95 0.80,1.14 1,444
Alone or with 1.17 0.88,1.56 509 0.97 0.70,1.34 428
others – spouse
Limiting long term illness
With others – with 1.08 0.89,1.30 615 1.23 ∗∗∗ 1.10,1.38 1,697
others
Change Alone – with others 2.16 ∗∗∗ 1.38,3.37 85 1.70 ∗∗∗ 1.34,2.17 312
Alone or with 1.06 0.85,1.32 509 0.85 0.67,1.07 428
others – spouse
Death 2001–2004
With others – with 0.97 0.75,1.26 615 1.23 ∗ 1.05,1.44 1,697
others
Change Alone – with others 1.25 0.70,2.22 85 1.31 0.96,1.79 312
Alone or with 0.72 0.50,1.04 509 0.61 0.36,1.04 428
others – spouse
N = 3,826 N = 10,519
a
Controlled for age, socio-economic status and LLTI in 1991. ∗ p<0.05∗∗ p<0.01∗∗∗ p<0.001.
Source: ONS LS, 1991–2004; authors’ analysis.
smaller than for those who only lived with others in 2001. Therefore, while health
selection may account for some of the association between living with others and
poor health, those living with others for longer time periods are also more likely
to change the rating of their health from good or fair to poor. It is worth noting,
however, when applying these results to the general population that only very small
proportions (two per cent of this subpopulation) moved to live with others and only
16% lived with others at both time points. Much larger proportions lived alone at
both time points.
Living Arrangement and Self-rated Health in Europe
Having established that older people living with others were generally more likely
to have poor self-rated health than those in other living arrangements in England
and Wales, at least in part due to health selection, the ESS was used to examine
this association in different regions of Europe. In the ESS, self-rated health was
measured using a five category variable including very good, good, fair, bad and
very bad health. We dichotomised the variable into very good, good and fair versus
bad and very bad self-rated health, with 23% of the sample rating their health as bad
or very bad. Note that this categorisation is in line with our LS and ELSA analysis
above.
Unmarried men living with others had the highest prevalence of poor self-rated
health in all regions, except the west region where the prevalence was higher for
100
80
60
40
20
0
North West South East North West South East
Men Women
Fig. 7.6 Proportion of the population aged 60+ with bad or very bad self-rated health by region
and gender, 19 European countries
(Source: ESS 2002 and 2004; authors’ analysis.)
those living alone (Fig. 7.6). Among unmarried women, the prevalence of poor self-
rated health was in general higher among those living alone than among those living
with others. Multivariate analyses, using logistic regression controlling for age and
socio-economic status demonstrated that, as for England and Wales, those living
with a spouse were less likely to have poor self-rated health than those living alone
(results not shown). However, results indicated that in all regions combined, women
living with others had a lower likelihood of poor self-rated health than those living
alone, in opposition to our findings from ELSA and the LS.
When we analysed regions separately, this association was significant for the
eastern region (p = 0.01) and was of borderline significance in the southern region
(p = 0.10), but there was no significant association in the northern and western
regions. There were also no significant associations among unmarried men for all
regions combined or by region. For self-rated health and mortality, we therefore
found differing results by region. While in England and Wales there was an associ-
ation between living with others and poor self-rated health, in southern and eastern
Europe, living alone was associated with poor self-rated health rather than living
with others.
Conclusions
What do these results show about the associations between living arrangements and
health and well-being? In England and Wales, as expected, older people living with
a spouse had the best levels of health and well-being, except for older women liv-
ing alone who rated their health as better than those living with a spouse. Among
the unmarried in England and Wales, those living alone rated their health bet-
ter than those living with others and were less likely to die during the follow-up
period, but they were also more likely to be depressed and lonely than those liv-
ing with others. We found, as expected, that these associations varied by region
of Europe, which is likely to be associated with differences in culture and welfare
regimes.
In northern Europe (here comprising Sweden, Norway, Finland and Denmark),
unlike all other regions, we found no association between living alone and low levels
of well-being. This may be due to the fact that older people in this region have pref-
erences for autonomy and independence, and receive generous benefits and support
from the state to enable them to live independently (Daatland, 1990; Daatland and
Herlofson, 2001; Millar and Warman, 1996; Reher, 1998). The association between
living alone and poorer emotional well-being in England and Wales was also found
in western, southern and eastern Europe, and may be associated with the impor-
tance of the role of the family in supporting older people in these regions. In south-
ern Europe, public policy reinforces the traditional focus of care and support by
the family (Twigg, 1996), and high percentages of the population in Mediterranean
Europe support the move of elderly parents into the child’s home when unable to
cope alone (Tomassini et al., 2004). In eastern Europe, there are still relatively low
levels of community social care (Botev, 1999; De Jong Gierveld et al., 2001; Tobis,
2000), and perhaps older people living alone are particularly disadvantaged and
unhappy due to a lack of family support. The UK is similar to Scandinavia in terms
of public attitudes to family responsibility (Katz et al., 2003), but welfare policy
places obligations on the nuclear family for support and care unlike Scandinavian
countries (Millar and Warman, 1996). Future research on this topic could include an
investigation of the further possible negative implications for well-being of living
alone for women in eastern, southern and western regions of Europe using larger
longitudinal datasets.
While we found that living with others was associated with poorer health in Eng-
land and Wales, in eastern and southern Europe, those living alone were likely to rate
their health as poorer than those living with others, in addition to a having a higher
likelihood of unhappiness. Self-rated health has a psychological component to it and
this finding may reflect a preference for living with others rather than alone in these
areas, especially in southern Europe. As already mentioned, lower levels of com-
munity based social services in eastern and southern Europe than in other regions
of Europe (Bond and Cabrero, 2007; Tobis, 2000) may mean that those living with
family receive higher levels of care and support than those living alone, resulting in
better self-rated health. Additionally, in the southern and eastern regions, it is more
normative to live with others and as there are higher proportions of people doing so,
they are less likely to be a group selected for poor health. In northern and western
Europe, where there are fewer older people living with others, those who do are
more likely to be in poor health. In line with this, we found that in England and
Wales, the association between living with others and poor self-rated health was at
least in part due to health selection. However, we also found that those living with
others for longer periods of time were more likely to develop an LLTI than those
living alone. It is important to note that in terms of generalising these results to the
population, the majority of unmarried older people live alone in England and Wales
and very few move to live with others. To fully eliminate the effects of health selec-
tion and to establish causal pathways, we would need longitudinal data with more
frequent follow ups. We would also need data on health status at the time of a living
arrangement transition, and more detailed measures of health status. This may be
possible for England once further waves of ELSA have been completed. European
regional differences in health selection could also be investigated further using lon-
gitudinal datasets such as SHARE when more rounds of data have been collected.
Some caution is needed when interpreting results because some analyses lacked
statistical power due to small sample sizes, especially of men living with others in
the northern and western regions. Further work using more detailed health indica-
tors might also reveal associations between health and disability and living arrange-
ments which we were unable to detect using the relatively crude indicators avail-
able in these datasets. Additionally, regional analysis may obscure country-level
associations and so more detailed country specific studies are also needed to elu-
cidate associations between living arrangements and well-being, and the factors
underlying them. Nevertheless, our findings demonstrate that whether or not living
alone has positive or negative consequences for older people’s health and emotional
well-being depends on the cultural context and the availability and acceptability of
alternative supports for those in need of assistance. With growing proportions of
the population in older age groups and a high propensity to live alone, numbers of
older people living alone are projected to increase in the coming decades in the UK
and throughout Europe. These increases may well have negative implications for the
health and well-being of some.
Acknowledgments We are grateful to the ONS for permission to use the LS, and for help pro-
vided by staff of the Centre for Longitudinal Study Information and User Support (CeLSIUS).
CeLSIUS is supported by the ESRC Census of Population Programme (Award Ref: RES-348-25-
0004). Census output is Crown copyright and is reproduced with the permission of the Controller
of HMSO and the Queen s Printer for Scotland. We are also grateful to the National Centre for
Social Research and the UK Data Archive for access to the ELSA data, and to the Norwegian
Social Science Data Service as the data archive and distributor of the ESS data. The authors alone
are responsible for the interpretation of the data.
References
Banks, J., Breeze, E., Lessof, C. and Nazroo, J. (2006). Retirement, Health and Relationships of
the Older Population in England: The 2004 English Longitudinal Study of Ageing (Wave 2),
Institute of Fiscal Studies, London.
Barnes, M., Blom, A., Cox, K. and Lessof, C. (2006) The Social Exclusion of Older People: Evi-
dence from the First Wave of the English Longitudinal Study of Ageing (ELSA), Social Exclu-
sion Unit, Office of the Deputy Prime Minister, London.
Bond, J. and Cabrero, G.R. (2007) Health and dependency in later life, in Bond, J., Peace, S.,
Dittmann-Kohli, F. and Westerhof, G. (eds.) Ageing in Society, Sage, London, pp. 113–141.
Botev, N. (1999) Older persons in countries with economies in transition, in Population Age-
ing, Challenges for Policies and Programmes in Developed and Developing Countries, United
Nations Population Fund and CBGS, New York and Brussels, pp. 85–100.
Braam, A.W., Prince, M.J., Beekman, T.F., Delespaul, P., Dewey, M.E., Geerlings, S.W., Kivela,
S.L., Lawlor, B.A., Magnusson, H., Meller, I., Peres, K., Reischies, F.M., Roelands, M.,
Schoevers, R.A., Saz, P., Skoog, I., Turrina, C., Versporten, A. and Copeland, J.R.M. (2005)
Physical health and depressive symptoms in older Europeans: Results from EURODEP, British
Journal of Psychiatry, 187: 35–42.
Brown, J.W., Liang, J., Krause, N., Akiyama, H., Sugisawa, H. and Fukaya, T. (2002) Transitions
in living arrangements among elders in Japan: Does health make a difference?, Journals of
Gerontology Series B-Psychological Sciences and Social Sciences, 57(4), S209–S220.
Burnette, D. and Mui, A.C. (1996) Psychological well-being among three successive cohorts of
older American women who live alone, Journal of Women & Aging, 8(1), 63–80.
Castro-Costa, E., Dewey, M., Steward, R., Banerjee, S., Huppert, F., Mendonca-Lima, C., Bula, C.,
Reisches, F., Wancata, J., Ritchie, K., Tsolaki, M., Mateos, R. and Prince, M. (2007) Prevalence
of depressive symptoms and syndromes in later life in ten European countries: The SHARE
study, British Journal of Psychiatry 191e Psychological Medicine, 6: 551–563.
Comstock, G.W. and Helsing, K. (1976) Symptoms of depression in two communities, Psycholog-
ical Medicine, 6: 551–563.
Crumley, B. (2003) Elder careless. Time Europe, London.
Daatland, S.O. (1990). What are families for? On family solidarity and preferences for help, Ageing
and Society, 10, 1–15.
Daatland, S.O. and Herlofson, K. (2001). Service systems and family care – substitution or comple-
mentarity? in Daatland, S.O. and Herlofson, K. (eds.) Ageing, Inter-generational Relationships,
Care Systems and Quality of Life. An Introduction to the OASIS Project, NOVA, Oslo,
pp. 53–61.
Daatland, S.O. and Herlofson, K. (2003) ‘Lost solidarity’ or ‘changed solidarity’: a comparative
European view of normative family solidarity, in Gender and Ageing Changing Roles and Rela-
tionships, Open University Press, Maidenhead, pp. 95–110.
De Jong Gierveld, J. (2003) Social networks and social well-being of older men and women living
alone, in Arber, S., Davidson, K. and Ginn, J. (eds.) Gender and Ageing Changing Roles and
Relationships, Open University Press, Maidenhead.
De Jong Gierveld, J., De Valk, H. and Blommesteijn, M. (2001) Living arrangements of older per-
sons and family support in more developed countries, Population Bulletin of the United Nations,
special issue, living arrangements of older persons: Critical Issues and Policy Responses,
42/43: 193–217.
Department of Health (2001) National Service Framework for Older People. Department of Health,
London.
Elman, C. and Uhlenberg, P. (1995) Early twentieth century co-residence of the elderly and their
children, Population Studies, 49: 501–517.
European Commission (2006) The Demographic Future of Europe – From Challenge to Opportu-
nity, European Commission, Luxembourg.
Fratiglioni, L., Wang, H.X., Ericsson, K., Maytan, M. and Winblad, B. (2000) Influence of social
network on occurrence of dementia: a community-based longitudinal study, Lancet, 355(9212):
1315–1319.
Garcia, E.L., Banegas, J.R., Perez-Regadera, A.G., Cabrera, R. and Rodriguez-Artalejo, F. (2005)
Social network and health-related quality of life in older adults: A population-based study in
Spain. Quality of Life Research, 14(2): 511–520.
Glaser, K., Murphy, M. and Grundy, E. (1997) Limiting long-term illness and household structure
among people aged 45 and over, Great Britain, 1991, Ageing and Society, 17: 3–20.
Glaser, K., Tomassini, C. and Grundy, E. (2004) Revisiting convergence and divergence: support
for older people in Europe, European Journal of Ageing, 1: 64–72.
Goldman, N., Korenman, S. and Weinstein, R. (1995) Marital status and health among the elderly,
Social Science & Medicine, 40: 1717–1730.
Grundy, E. (1989) Longitudinal perspectives on the living arrangements of the elderly, in Jefferys,
M. (ed.) Growing Old in the Twentieth Century, Routledge, London, pp. 128–150.
Grundy, E. (1993) Moves into supported private households among elderly people in England and
Wales, Environment and Planning A, 25: 1467–1479.
Grundy, E. (1996) Population review: The population aged 60 and over, Population Trends, 84:
14–20.
Grundy, E. (1999) Changing role of the family and community in providing support for the elderly,
in Cliquet, R. and Nizamuddin, M. (eds.) Population Ageing. Challenges for Policies and Pro-
grammes in Developed and Developing Countries, UNFPA, New York, pp. 103–122.
Grundy, E. (2000) Co-residence of mid-life children with their elderly parents in England and
Wales, changes between 1981 and 1991, Population Studies, 54, 193–206.
Grundy, E. (2001) Living arrangements and the health of older persons in developed countries,
Population Bulletin of the United Nations, special issue, living arrangements of older persons:
Critical Issues and Policy Responses, 42/43: 311–329.
Grundy, E. (2006) Ageing and vulnerable elderly people: European perspectives, Ageing and Soci-
ety, 26: 105–134.
Grundy, E. and Sloggett, A. (2003) Health inequalities in the older population: the role of personal
capital, social resources and socio-economic circumstances, Social Science & Medicine, 56:
935–947.
Gustavson, K. and Lee, C. (2004). Alone and content: frail seniors living in their own homes
compared to those who live with others, Women and Aging, 16(3–4): 3–18.
Hahn, B.A. (1993) Marital status and women’s health, the effect of economic and marital acquisi-
tions, Journal of Marriage and the Family, 55: 495–504.
Hall, M. and Havens, B. (2001) The Effects of Social Isolation and Loneliness on the Health of
Older Women, Centres of Excellence for Women’s Health, Winnipeg.
Hattersley, L. and Creeser, R. (1995) Longitudinal Study 1971–1991. History, Organisation and
Quality of Data, ONS, London.
Hughes, M.E. and Waite, L.J. (2002) Health in household context: living arrangements and health
in late middle age, Journal of Health and Social Behavior, 43(1): 1–21.
Hughes, M.E., Waite, L.J., Hawkley, L.C. and Cacioppo, J.T. (2004) A short scale for measuring
loneliness in large surveys: results from two population-based studies, Research on Aging, 26:
655–672.
Iacovou, M. (2000) The living arrangements of elderly Europeans, No. 2000–2009, Institute for
Social and Economic Research, Essex.
Idler, E.L. and Benyamini, Y. (1997) Self-rated health and mortality: a review of twenty-seven
community studies, Journal of Health and Social Behavior, 38: 21–37.
Iliffe, S., Tai, S.S., Haines, A., Gallivan, S., Goldenberg, E., Booroff, A. and Morgan, P. (1992)
Are elderly people living alone an at risk group? British Medical Journal, 305: 1001–04.
Iwata, N. and Buka, S. (2002) Race/ethnicity and depressive symptoms: a cross-cultural/ethnic
comparison among university students in East Asia, North and South America, Social Science
& Medicine, 55: 2243–2252.
Josselson, R. (1996) The Space Between Us. Exploring the Dimensions of Human Relationships,
Sage, Thousand Oaks, CA.
Jowell, R. (2003) European Social Survey 2002/2003: Technical Report, Centre for Comparative
Social Surveys, City University, London.
Katz, R., Daatland, S.O., Lowenstein, A., Bazo, M.T., Ancizu, I., Herlofson, K., Mehlhausen-
Hassoen, D. and Prilutzky, D. (2003) Family norms and preferences in intergenerational rela-
tions, in Bengston V. and Lowenstein, A. (eds.) Global Ageing and Challenges to Families,
Aldine de Grutyer, New York, pp. 305–326.
Kendig, H., Dykstra, P.A., van Gaalen, R.I. and Melkas, T. (2007) Health of aging parents and
childless individuals, Journal of Family Issues, 28(11): 1457–1486.
Kharicha, K., Iliffe, S., Harari, D., Swift, C., Gillmann, G. and Stuck, A. (2007) Health risk
appraisal in older people 1: are older people living alone an at-risk group? British Journal
of General Practice, 57: 271–276.
Klinenberg, E. (2002). Heat Wave: A Social Autopsy of Disaster in Chicago, University of Chicago
Press, Chicago.
Kohler, H.P., Behrman, J.R. and Skytth, A. (2005) Partner + Children=Happiness? The
effects of partnerships and fertility on well-being, Population and Development Review, 31:
407–455.
Kohli, M. and Albertini, M. (2006) The Impact of Welfare and Family Regimes on Transfers
Between Older Parents and Their Adult Children, Varenna, Italy.
Koskinen, S., Joutsenniemi, K., Martelin, T. and Martikainen, P. (2007) Mortality differ-
ences according to living arrangement, International Journal of Epidemiology, 36(6):
1255–1264.
Lee, G.R., DeMaris, A., Bavin, S. and Sullivan, R. (2001) Gender differences in the depressive
effect of widowhood in later life, Journals of Gerontology Series B-Psychological Sciences
and Social Sciences, 56(1): S56–61.
Lehtinen, V., Sohlman, B. and Kovess-Masfety, V. (2005) Level of positive mental health in the
European Union: results from the Eurobarometer 2002 Survey, Clinical Practice and Epidemi-
ology in Mental Health, 1: 9.
Lund, R., Modvig, J., Due, P. and Holstein, B.E. (2000) Stability and change in structural social
relations as predictor of mortality among elderly women and men, European Journal of Epi-
demiology, 16(12): 1087–1097.
Marmot, M., Banks, J., Blundell, R., Lessof, C. and Nazroo, J. (2002). Health, Wealth and
Lifestyles of the Older Population in England: The 2002 English Longitudinal Study of Ageing,
Institute for Fiscal Studies, London.
McGarry, K. and Schoeni, R.F. (2000) Social security, economic growth, and the rise in elderly
widows independence in the twentieth century, Demography, 37(2): 221–236.
Michael, Berkman, L.F., Colditz, G.A. and Kawachi, I. (2001) Living arrangements, social inte-
gration, and change in functional health status, American Journal of Epidemiology, 153(2):
123–131.
Michael, R., Fuchs, V. and Scott, S. (1980) Changes in the propensity to live alone, Demography,
19: 39–53.
Millar, J. and Warman, A. (1996) Family Obligations in Europe, Family Policy Studies Centre,
London.
Mindel, C.H. and Wright, R. (1982). Differential living arrangements among the elderly and their
subjective well-being, Activities, Adaptation and Aging, 2: 25–34.
Montgomery, S.M., Netuveli, G., Hildon, Z. and Blane, D. (2007) Does financial disadvantage at
older ages eliminate the potential for better health?, Journal of Epidemiology and Community
Health, 61: 891–895.
Mor, V., Murphy, J., Masterson-Allen, S., Willey, C., Razmpour, A., Jackson, M.E., Greer, D. and
Katz, S. (1989) Risk of functional decline among well elders, Journal of Clinical Epidemiology,
42(9): 895–904.
Mui, A.C. and Burnette, D. (1994) A comparative profile of frail elderly persons living alone and
those living with others, Journal of Gerontological Social Work, 21: 5–26.
Murata, C., Kondo, T., Hori, Y., Miyao, D., Tamakoshi, K., Yatsuya, H., Sakakibara, H.
and Toyoshima, H. (2005) Effects of social relationships on mortality among the elderly
in a Japanese rural area: An 88-month follow-up study, Journal of Epidemiology, 15(3):
78–84.
Murphy, M., Grundy, E. and Kalogirou, S. (2007) The increase in marital status differences in
mortality up to the oldest age in seven European countries, 1990–99, Population Studies, 61(3):
287–298.
Netuveli, G., Wiggins, R.D., Hildon, Z., Montgomery, S.M. and Blane, D. (2006) Quality of life
at older ages: evidence from the English Longitudinal Study of Ageing (wave 1), Journal of
Epidemiology Community Health, 60: 357–363.
ODPM. (2006) The Social Exclusion of Older People: Evidence from the First Wave of the English
Longitudinal Study of Ageing (ELSA), HMSO, London.
Office for National Statistics (2004) Census 2001 Definitions, Office for National Statistics,
London.
Olafsson, S. (2003) Welfare trends of the 1990s in Iceland, Scandinavian Journal of Public Health,
31: 401–404.
Pampel, F.C. (1992) Trends in living alone among the elderly in Europe, in Rogers, A. (ed.) Elderly
Migration and Population Redistribution, Belhaven Press, London, pp. 97–117.
Pillemar, K., Moen, P., Wethington, E. and Glasgow, N. (2000) Social Integration in the Second
Half of Life, Johns Hopkins University Press, Baltimore.
Pizzetti, P., Manfredini, M. and Lucchetti, E. (2005) Variations in late-age mortality by household
structure and marital status in Parma, Italy, Ageing and Society, 25: 305–318.
Radloff, L.S. (1977) The CES-D scale: A self-report depression scale for research in the general
population, Applied Psychological Measurement, 1: 385–401.
Reher, D.S. (1998) Family ties in Western Europe: persistent contrasts, Population Development
Review, 24(2): 203–234.
Russell, D.W. (1996) UCLA Loneliness scale (version 3): reliability, validity, and factor structure,
Journal of Personality Assessment, 66: 20–40.
Sarwari, A.R., Fredman, L., Langenberg, P. and Magaziner, J. (1998) Prospective study on the
relation between living arrangement and change in functional health status of elderly women,
American Journal of Epidemiology, 147(4): 370–378.
Silverstein, M. and Bengston, V. (1994) Does intergenerational social support influence the psy-
chological well-being of older parents? The contingencies of declining health and widowhood,
Social Science & Medicine, 38(7): 943–957.
Soule, A., Babb, P., Evandrou, M., Balchin, S. and Zealey, L. (2005) Focus on Older People, Office
for National Statistics, London.
Steffick, D.E. (2000) HRS/AHEAD Documentation Report – Documentation of Affective Func-
tioning Measures in the Health and Retirement Study, University of Michigan, Ann
Arbour.
Sundstrom, G. and Tortosa, M.A. (1999). The effects of rationing home-help services in Spain and
Sweden: a comparative analysis, Aging and Society, 19: 343–361.
Tobis, D. (2000) Moving from Residential Institutions to Community-Based Social Services in Cen-
tral and Eastern Europe and the Former Soviet Union, The International Bank for Reconstruc-
tion and Development, The World Bank, Washington D.C.
Tomassini, C., Glaser, K., Wolf, D., van Groenou, M. and Grundy, E. (2004) Living arrangements
among older people: an overview of trends in Europe and the USA, Population Trends, 115:
24–34.
Twigg, J. (1996) Issues in informal care, in OECD (ed.) Caring for Frail Elderly People: Policies
in Evolution, Vol. Social Policy Studies No. 19, OECD, Paris, pp. 81–92.
Ulbrich, P.M. and Warheit, G.J. (1989) Social support, stress and psychological distress among
older black and white adults, Journal of Aging and Health, 1: 286–305.
United Nations (2002) Report of the Second World Assembly on Ageing Madrid, 8–12 April
2002,United Nations, New York.
Walter-Ginzburg, A., Blumstein, T., Chetrit, A. and Modan, B. (2002) Social factors and mortality
in the old-old in Israel: The CALAS study, Journals of Gerontology Series B-Psychological
Sciences and Social Sciences, 57(5): S308–S318.
Wenger, G.C. (1984) The Supportive Network: Coping with Old Age, Allen and Unwin, London.
Wilcox, S., Evenson, K.R., Aragaki, A., Wassertheil-Smoller, S., Mouton, C.P. and Loevinger, B.L.
(2003) The effects of widowhood on physical and mental health, health behaviors, and health
outcomes: The Women’s Health Initiative, Health Psychology, 22(5): 513–522.
Wilson, K.C.M., Chen, R., Taylor, S., McCracken, C.F.M. and Copeland, J.R.M. (1999) Socio-
economic deprivation and the prevalence and prediction of depression in older community res-
idents, British Journal of Psychiatry, 176: 549–553.
Zunzunegui, M.V., Beland, F. and Otero, A. (2001) Support from children, living arrangements,
self-rated health and depressive symptoms of older people in Spain, International Journal of
Epidemiology, 30(5): 1090–1099.
Chapter 8
Stepparenting and Mental Health
Peteke Feijten, Paul Boyle, Zhiqiang Feng, Vernon Gayle

and Elspeth Graham
Introduction
Demographic changes in the Western world over the last few decades, such as later
marriages, lower fertility, increasing divorce rates and rising rates of cohabitation,
have brought about significant changes in household formation and composition.
One outcome is a growing number of stepfamilies, where a parent, whether never
married, separated, widowed or divorced, forms a new marriage or partnership.
Nowadays, most stepfamilies result from divorce, while in the past they were more
likely to result from widowhood.
Despite the rising incidence of stepfamilies and the demographic and social
differences between stepfamilies and traditional families with two biological par-
ents, researchers concerned with family life and parenting were relatively slow
to acknowledge the importance of such non-traditional families (Ferri and Smith,
1998; Utting, 1995). While there has been an increase in social science research on
stepfamilies in recent years much remains to be done (Coleman et al., 2000). For
example, numerous researchers have explored the effect of living in a stepfamily
on children, including studies of their psychological well-being but, perhaps sur-
prisingly, little research has considered the potential psychological impacts on the
adults (stepparents and their partners) of living in such a household arrangement.
This is the focus of this study.
Background
In the UK most people live in ‘traditional’ households, but non-traditional house-

holds are becoming more common. Of those marrying during the 1990s, nearly 50%
will end up divorced if current trends continue (Allan, 1999), and a growing num-
ber of divorcees are starting new relationships. Haskey (1994) estimated that 12%
of British children will live in a stepfamily before their sixteenth birthday. Recent
P. Boyle (B)
School of Geography and Geosciences, University of St Andrews, St Andrews, KY16 9AL, UK
e-mail: p.boyle@st-andrews.ac.uk

152 P. Feijten et al.
estimates for Britain suggest that about 40% of mothers will experience being a lone
parent and about 75% of lone mothers will go on to form a stepfamily (Ermisch and
Francesconi, 2000). Currently nearly 90% of stepfamilies involve children living
with their mother and a new male partner (Finch, 2002). Importantly, stepfamilies
differ from traditional families demographically because they tend to include more
and older children than do first families (Haskey, 1994). The greater complexity
of intra-household relationships in stepfamilies provides scope for tensions to arise
and, perhaps, increases the potential for negative health impacts for both the children
and the parents.
Numerous studies highlight the strains that every-day life in stepfamilies may
entail and the effects this may have on stepchildren’s health and well-being (Brown
and Booth, 1996; Pryor and Rodgers, 2001). Not only do stepchildren experience the
breakdown of their parents’ relationship but they often feel relatively neglected by
the biological parent. This may be combined with the potentially disruptive effects
of having to divide their time between two homes. Most studies of the effects of
remarriage on children fail to show a benefit, despite the financial advantages that
usually result (Fergusson et al., 1994; Pagani et al., 1998; Walper, 1995; Duncan and
Hoffman, 1985; Zill, 1988). Some findings point to negative effects, with stepchil-
dren performing worse at school (Pong, 1997; Teachman et al., 1996), five-year old
children in stepfamilies being significantly more at risk of behavioural and devel-
opmental problems than children in traditional families (Wadsworth et al., 1985),
and a higher risk of drinking alcohol, drug abuse and problem behaviour among
schoolchildren living with a stepfather (Del Carmen et al., 2002; Mekos et al., 1996).
While Joshi et al. (1999) found that maternal educational attainment and, to a lesser
extent, family economic circumstances eliminated the relationship between family
structure and children’s cognitive and behavioural outcomes, the majority of quan-
titative studies suggest that stepchildren are at greater risk of a range of problems
(Coleman et al., 2000; Ram and Hou, 2003).
There is also a considerable literature on the effects of family arrangements and
marital status on adult health. Higher mortality rates among the unmarried, those
who live alone and the divorced, compared to those who are married, are well estab-
lished (Seeman et al., 1987; Trovato and Lauris, 1989; Gardner and Oswald, 2004).
More depressive symptoms are apparent among both the recently separated (Neff
and Schluter, 1993) and those who have been separated for longer (Richards et al.,
1997), even when mental health status prior to separation/divorce is taken into
account (Wade and Pevalin, 2004). In some studies, the beneficial effects of mar-
riage are found for men but not women (Berkman and Syme, 1979; Avlund et al.,
1998). Lone parenthood has also been studied in some detail. Hope et al. (1999)
found that lone parents suffer higher levels of mental distress than other parents,
although this may be related to the significantly higher poverty levels they experi-
ence (Keirnan and Mueller, 1998; Shouls et al., 1999). Notably, within this large
literature, there are virtually no studies examining explicitly the effects of living in
a stepfamily on stepparents’ and their partner’s health.
Another strand of the literature which is relevant to this study considers ‘mar-
ital quality’ and how people cope with the redefinition of kinship that follows
8 Stepparenting and Mental Health 153
divorce and remarriage (overviews are provided by Coleman and Ganong, 1990;
Pasley et al., 1993). Remarriage has been found to lead to lower marital quality and
satisfaction than first marriage (Clingempeel, 1981; White and Booth, 1985), and
this may have an impact upon mental health. In a qualitative study by Simpson
(1994), it was argued that the roles of remarried persons compared to married per-
sons, and stepparents compared to biological parents, are less well defined and thus
harder to fulfil with confidence and satisfaction. This is sometimes labelled as the
‘incomplete institution hypothesis’ (Cherlin, 1978). We found one study specifi-
cally focusing on the experience of stepparenting and how this affected marital
quality and the stepparent-stepchild relationship. Ambert (1986) showed that an
intimate relationship between stepparents and stepchildren was harder to establish
when stepchildren were not resident and that the birth of a common child into the
stepfamily made the stepparent-stepchild relationship better for men, but not for
women.
Thus, stepfamily arrangements are potentially stressful since they involve the
negotiation of different intra-household relationships that may introduce new
sources of tension. For example, Hetherington and Jodl (1994) found that step-
parents remain less engaged and more authoritarian in parenting stepchildren than
in parenting their biological children, illustrating the scope for friction between
biological parents and stepparents. Yet a recent wide-ranging literature review on
remarriage and stepparenting cited only a small number of studies on the psycholog-
ical health of adults, the majority of which examined the general effects of remar-
riage (Coleman et al., 2000). A study by Ferri and Smith (1998) suggested that
adults in stepfamilies were more likely to express ‘negative feelings’ and suffer
from depression than those in first families. Also, a recent study considered the rela-
tionship between depression and being a parent in the US (Evenson and Simon,
2005). Parents with young children living at home were shown to have significantly
higher rates of depression than non-parents and those with adult children who have
left home. However, no increased risk of depression for stepparents with minor aged
stepchildren was found and this study only considered the mental health of steppar-
ents, not their partners.
In addition to the lack of previous work on the topic, another challenge to any
quantitative study of stepparenting comes from relatively recent, largely qualitative
studies that have begun to undermine the idée fixe that divorce is always harmful,
at least to children (Smart, 2003). These new approaches benefit from life course
perspectives and deploy different conceptual categories, which recognise the wide
variety of stepfamilies that exist. In developing our quantitative research design, we
have tried to incorporate a greater subtlety into our analysis than some previous
quantitative studies, by recognising that there are many types of stepfamily and that
health impacts may change over time.
Coleman et al. (2000) state that more longitudinal quantitative studies of the
effects of stepparenting are required. One particularly important reason for this is
that cross-sectional studies cannot control adequately for selection effects. Thus,
while stepparenting may result in poorer mental health, the opposite effect is also
possible; those prone to poorer mental health may be more likely to end up living
in stepfamilies (Amato, 2000). Only with longitudinal data could this be explored.
The National Child Development Study (NCDS) collects data for a large birth cohort
based on all children born in a single week in 1958 and provides information from
birth throughout childhood and adolescence into young and middle age adulthood.
These data made it possible to control for adolescent characteristics which may
influence subsequent mental health. This is a simple but effective way of providing
improved statistical control for the potential of increased representation of mentally
ill people into step-parenthood.
In sum, there is a dearth of research on the mental health of adults living in
stepfamilies, a need for a large scale quantitative longitudinal study in the UK and
a demand that such a study gives due consideration to new and challenging con-
ceptualisations of family arrangements. Our study, a longitudinal analysis using
secondary data from a British birth cohort study, is designed to respond to all three
of these points.
Research Aims and Hypotheses
The study aimed to fill the gap in knowledge about the relationship between step-
parenting and mental health in the UK. Our main research question was, ‘how does
the mental health of adults in stepfamilies differ from the mental health of adults in
otherwise comparable first families?’ We were also interested in several dimensions
of stepfamilies that are identified in the literature as affecting marital quality and
personal well-being. These relate to the characteristics of the stepparent, such
as age, income and attitude, and the characteristics of the children in the stepfam-
ily, such as age and whether they are resident or non-resident. Thus, we derived the
following hypotheses:
1. Stepparents have worse mental health outcomes than parents in first families.
2. The partners of stepparents have worse mental health outcomes than parents in
first families.
These first two hypotheses are at the heart of our study, as described above.
3. The poorer the stepfamily, the greater the negative effect on mental health for
both partners.
This has been identified as an understudied area in stepfamily research, but
we base our hypothesis on reports that financial issues are one of the primary
sources of stress in stepfamilies (Coleman and Ganong, 1990).
4. The younger the stepparent, the greater the negative effect on mental health for
both partners.
Firstly, young age indicates a lack of experience with children. Palisi et al.
(1991) found that stepparents who had previous parenting experience performed
better at stepparenting than those with no experience. Secondly, young age can be
interpreted as a proxy for time spent in the stepfamily. There is some evidence
that the family situation improves when the stepparent is longer in the family
(Pasley et al., 1993).
5. The presence of a child born to the two partners in a stepfamily reduces the
likelihood of poor mental health outcomes for both partners.
Studies in the past (Berman, 1980; Ambert, 1986) have found that a common
child increases the quality of the relationship of partners in a stepfamily, and this
leads us to expect that it also positively affects the mental health of the partners.
However, the quality of the relationship between stepparents and stepchildren
has not been found to improve as a result of the birth of a common child (White
and Booth, 1985).
6. The presence of adolescent stepchildren increases the likelihood of poor mental
health outcomes for both partners in stepfamilies.
In a study on the well-being of stepchildren, Hetherington and Clingempeel
(1992) showed that stepparents found it particularly hard to cope with adoles-
cent stepchildren who were often coercive and hostile towards the stepparent,
whereas with younger children it was often easier for the stepparent to find a
parenting style that worked. The relationship between stepfathers and adolescent
stepdaughters was found to be especially problematic.
7. The likelihood of poor mental health outcomes for both parents in stepfamilies
increases where stepparents hold more traditional views.
Those with traditional views often have an authoritative parenting style,
and the literature shows that authoritative parenting styles do not work well
with stepchildren (especially older stepchildren) (Hetherington et al., 1988;
Hetherington, 1991). Therefore, we expect traditional views to lead to an
increased risk of poor mental health for such parents.
8. The presence of non-resident children in the household increases the likelihood
of poor mental health outcomes for both partners in stepfamilies.
Relationships between stepparents and resident children have been shown to
be closer than between stepparents and non-resident stepchildren. Stepparents
with non-resident stepchildren may have to manage a relationship with the ex-
spouse and it may also be difficult to develop close bonds with stepchildren who
rarely visit (Ambert, 1986). These issues may add stress to the relationship and
subsequently affect the stepparent’s mental health.
Methods
The data used for this study is the National Child Development Study (NCDS),
which is a sample of all babies born in the UK in one week in the spring of 1958.
The sample originally included 17,416 respondents and they have been returned
to seven times to collect a variety of data on issues including: mental health, part-
nership histories, and other time-invariant and time-varying demographic, health
and socio-economic variables (University of London, 2007). We focused in this
study on characteristics as recorded in 1991 when the sample members were aged
33, but we also included some variables from earlier in the life course which
we expected to be related to subsequent mental health status (University of Lon-
don, 2000). The sample (n = 6,121) includes those with children, those with valid
measures of behavioural/mental problems at age 16 and age 33, and those with
family status information. Descriptive statistics for all the variables are given in
Table 8.1.
Our outcome variable of interest is mental health. The measurement instrument
used to measure mental health in the NCDS is the Malaise Inventory Scale (MIS)
developed by Rutter et al. (1970). It consists of 24 questions which are designed
to capture depression and anxieties, obsessions and phobias. It has commonly been
Table 8.1 Descriptive statistics for variables used in the study
Variable Number Percentage
Mental health age 33

Good health 5395 88.1
Poor health 726 11.9
Behavioural problems at age 16
Home Behaviour Scale score < 7 4823 78.8
Home Behaviour Scale score >= 7 1298 21.2
Family type age 33
First family 4638 75.8
Lone parent 420 6.9
Stepparent 400 6.5
Partner of stepparent 494 8.1
Both stepparents 169 2.8
Sex
Male 2,714 44.3
Female 3,407 55.7
Employment status
Full-time working 3,264 53.3
Part-time working 1,356 22.2
Unemployed 195 3.2
Other not working 1,306 21.3
Highest completed educational level
No education completed 1,076 17.6
CSE 2-5 1,201 19.6
O levels 2,482 40.5
A levels 604 9.9
Degree 684 11.2
Unknown 74 1.2
Social class
Unskilled 280 4.6
Partly skilled 1,046 17.1
Skilled manual 1,137 18.6
Skilled non-manual 1,388 22.7
Managerial / technical 1,642 26.8
Professional 212 3.5
Unknown 416 6.8
Domestic tension age 7
No 5,194 84.9
Yes 255 4.2
Unknown 672 11.0
Variable Number Percentage
School abilities age 7

Most abilities below average 4,181 68.3
Most abilities on or above average 1,331 21.7
Unknown 609 9.9
Financial hardship age 7
No 5,071 82.8
Yes 378 6.2
Unknown 672 11.0
Living with natural mother age 16
No 245 4.0
Yes 5,876 96.0
Father interested in child’s education age 16
No 781 12.8
Yes 3,022 49.4
Unknown 2,318 37.9
School abilities at age 16
More than half below CSE 1/O levels 1,189 19.4
Half or more above CSE 1/O levels 1,989 32.5
Unknown 2,943 48.1
Variable Mean S.d.
Mean number of children by family type
Overall mean 2.1 1.0
First family 2.0 0.8
Lone parent 1.9 1.0
Stepparent 2.7 1.2
Partner of stepparent 2.5 1.1
Both stepparents 4.2 1.5
Source: Authors’ analysis based on survey data.
used as a mental health screening instrument and several studies have tested and
confirmed the alpha reliability and internal consistency of this scale (e.g., Cherlin
et al., 1998; Hirst and Bradshaw, 1983). The distribution of the MIS scale is highly
skewed, and a common solution to deal with this is to divide the scale into a binary
variable. As in other studies, such as Flouri and Buchanan (2002) and Power et al.
(1988), we defined a score of 7 or higher out of 24 as an indicator of poor men-
tal health. As Chase-Landsdale et al. (1995: 1619) point out, the MIS is merely a
screening instrument, and a score of 7 or higher must be interpreted as ‘a clinical
cut off score, indicating a high likelihood of the presence of mental illness and the
need for psychiatric help’. Twelve percent of adults aged 33 have a score over 6 on
the MIS.
We also wanted to control for proneness to poor mental health at the onset of
adulthood, because we hypothesised that this may influence whether people are
more or less likely to end up living in a stepfamily – a potential sample selec-
tion effect. This was operationalised through the Home Behaviour Scale (HBS),
also developed by Rutter et al. (1970). This measure was collected in 1974 when
respondents were aged 16. The instrument consists of a 22-item scale that is meant
to signal both externalizing (where the child shows under-controlled behaviour
such as aggression or disobedience) and internalizing (where the child shows
over-controlled behaviour such as anxiety or depression) behavioural disorders in
children. Again, a score of 7 or higher was chosen as a threshold for defining
behavioural problems. Twenty one per cent of children at age 16 have a score over
6 on the HBS.
In order to distinguish different types of (step)families, we created a categorical
variable with five categories: (1) respondents in first families, (2) respondents in
lone parent families, (3) respondents who are stepparents, (4) respondents who are
the partners of stepparents and (5) respondents who are stepparents and partners
of stepparents at the same time (i.e., both partners have children from a previous
relationship). We label this group ‘dual stepfamilies’, and the results show that this
group is particularly important (not all studies consider this rarer group).
Finally, we extracted additional explanatory variables expected to influence men-
tal health. We used variables from different points in the life course. Some reflect
the status of the sample respondents in the year of analysis, 1991, when respondents
were 33 years old. These included the sex of the respondent, their economic status,
highest educational qualification and social class. We also included the number of
children in the family, as stepfamilies are larger than first families on average. Then,
we included variables about respondents’ characteristics at ages 7 and 16, to control
for previous circumstances expected to be associated with mental health outcomes
in later life (see Cherlin et al., 1998; Flouri and Buchanan, 2002). These included
domestic tension at age 7, financial hardship in the child’s household at age 7,
whether the child lived with his/her natural mother at age 16, whether the father was
interested in the child’s education at age 16, and school abilities at ages 7 and 16.
The method used is logistic regression, with mental health status at age 33
(1991) as the binary dependent variable (0 = non-poor mental health; 1 = poor men-
tal health). All models include the control variables described above, and the results
are expressed as parameter estimates in the tables. We first estimated a model with-
out behavioural problems in adolescence, and then a model that included this vari-
able, where we were particularly interested in the effects of the interaction between
behavioural problems at age 16 and family status at age 33 on mental health sta-
tus at age 33. We also graphed the coefficients by family status. In order to better
facilitate appropriate visual comparisons between categories of the explanatory vari-
able, we have plotted comparison intervals as suggested by Firth (2003) and further
illustrated by Gayle and Lambert (2007).
Results
Description of Stepfamilies
We distinguish between stepparents, partners of stepparents, and families

where both partners are stepparents to each other’s children in our study
Table 8.2 Demographic characteristics by family type
Average Average Average age

number of number of range
Distribution children children (incl. children
family type (resident non-resident (resident Average age
(%) children only) children) children only) (step)parent∗
First family 75.6 2.0 2.0 2.8 33.3

Lone parent family 7.2 1.7 1.9 3.0 –
All stepfamilies 17.3 1.8 2.8 4.5 34.7
Stepparent 6.4 1.7 2.7 4.1 37.5
Partner of stepparent 8.1 1.8 2.5 4.9 32.0
Both stepparents 2.8 2.3 4.2 4.6 36.1
Source: Authors’ analysis based on NCDS sweep 1991.

∗
This statistic is based on partners of NCDS cohort members, because cohort members themselves
were all aged 33 at the time of the survey.
(‘dual stepfamilies’). The distribution of respondents over these family types is

shown in Table 8.2 (column 1). Table 8.2 also shows that adults living in stepfami-
lies are different from adults living in first families and lone parent families, in terms
of age, number of children in the household and the age range of these children (see
also Haskey, 1994). The second and third columns show the average number of chil-
dren per family type (excluding and including non-resident children). Columns four
and five show the average age range of the children in the household, and the aver-
age age of (step)parents per family type. It can be seen that adults in stepfamilies are
on average older and have more children, who are of a wider age range, than people
in first families and lone parent families.
Association Between Being a (Partner of a) Stepparent

and Mental Health
Table 8.3 shows the numbers and percentages of people with poor and non-poor
mental health, by family type, at age 33. Adults in stepfamilies (both stepparents
and their partners) were more likely to suffer poor mental health than adults in first
families, but less likely than lone parents.
Table 8.3 Poor mental health by family type
Non-poor mental health Poor mental health

n % N %
First family 5,631 90.7 575 9.3

Lone parent family 434 73.9 153 26.1
Stepfamily 176 83.1 240 16.9
Source: Authors’ analysis based on NCDS sweep 1991.

Health Selection into Stepfamilies?

As argued above, it is possible that part of this association is due to health selection
into stepfamilies. Table 8.4 shows the numbers and percentages of people with and
without behavioural problems at age 16 who ended up in a stepfamily at age 33.
Those with behavioural problems at age 16 were significantly more likely to live
in a stepfamily as an adult (22.9%) than those who had no behavioural problems at
age 16 (15.9%). This supports the idea that at least part of the worse mental health
of (partners of) stepparents is due to the selection of people prone to poorer mental
health into stepfamilies. This is therefore accounted for in the models below.
Table 8.4 Behavioural problems at age 16 by stepfamily status at age 33
In a stepfamily at age 33?

No Yes Total
Behavioural problems
at age 16? n % n % %
No 4,057 84.1 766 15.9 100

Yes 1,001 77.1 297 22.9 100
Pearson chi2(1) = 34.9; p < 0.000.

Source: Authors’ analysis based on NCDS sweeps 1974 and 1991.
Modelling Results
All hypotheses were tested in models with and without control variables, which
included: sex, household size, employment status, highest completed education,
social class, domestic tension at age 7, financial hardship at age 7, school abilities
at ages 7 and 16, living with the natural mother at age 16, and father interested
in education at age 16. Table 8.5 shows the results from two multivariate anal-
yses. Model 1 does not take behavioural problems at age 16 into account while
Model 2 does.
We see, first of all, that the number of children in the household does not affect
the risk of having poor mental health (Model 1). Gender does have an influence:
our result replicates the well-known finding that women have a higher risk of poor
mental health than men (Bird and Fremont 1991). The employment status variable
shows that people who do not work are at increased risk of having poor mental
health, especially those who are unemployed. Also, those with lower levels of edu-
cation and from lower social classes have an increased risk of poor mental health.
Childhood circumstances at age 7 also affect mental health at age 33. Living
in a household with domestic tension or financial hardship, or performing below
average at school increases the risk of being in poor mental health. Those who lived
with their natural mother or had a father who was interested in their education at age
16 have better mental health at age 33. Also school abilities at age 16 are positively
associated with mental health at age 33. Our main variable of interest at age 16 is
whether the respondent had behavioural problems. Model 2 shows that those with
Table 8.5 Modelling results for mental health status at age 33 (1991) (n = 6,121)
Model 1 Model 2
Variable Coef. p-value Coef. p-value
Constant –2.378 0.000 –2.470 0.000

Behavioural problems age 16
HBS score < 7 0.000
HBS score >= 7 0.554 0.000
Family type age 33
First family 0.000 0.000
Lone parent family 0.692 0.000 0.572 0.001
Stepparent 0.401 0.017 0.282 0.187
Partner of stepparent 0.329 0.017 0.211 0.240
Dual stepfamily 0.869 0.000 0.966 0.000
Behavioural problems age 16 ∗ Family type age 33
First family 0.000
Lone parent family 0.194 0.468
Stepparent 0.257 0.452
Partner of stepparent 0.206 0.463
Dual stepfamily –0.320 0.429
Number of children in the household (incl. non-resident ) –0.011 0.797 –0.021 0.630
Sex
Male 0.000 0.000
Female 0.658 0.000 0.624 0.000
Employment status age 33
Fulltime working 0.000 0.000
Part-time working –0.080 0.556 –0.076 0.576
Unemployed 0.628 0.001 0.572 0.003
Other not working 0.228 0.079 0.224 0.086
Highest completed educational level age 33
No education completed 0.000 0.000
CSE levels 2–5 –0.410 0.001 –0.373 0.002
O levels –0.523 0.000 –0.481 0.000
A levels –0.734 0.000 –0.663 0.001
Degree or subdegree –1.064 0.000 –0.985 0.000
Unknown –0.617 0.119 –0.528 0.184
Social class age 33
Unskilled 0.000 0.000
Partly skilled –0.157 0.371 –0.139 0.434
Skilled manual –0.235 0.212 –0.219 0.250
Skilled non-manual –0.431 0.018 –0.401 0.029
Managerial / technical –0.510 0.008 –0.469 0.016
Professional –0.802 0.063 –0.725 0.094
Unknown 0.195 0.202 0.221 0.278
Domestic tension age 7
No 0.000 0.000
Yes 0.497 0.004 0.428 0.015
Unknown –1.449 0.152 –1.454 0.158
Financial hardship age 7
No 0.000 0.000
Model 1 Model 2
Variable Coef. p-value Coef. p-value
Yes 0.370 0.012 0.360 0.016

Unknown 1.474 0.141 1.471 0.149
School abilities age 7
Most abilities on or above average 0.000 0.000
Most abilities below average 0.231 0.023 0.215 0.037
Unknown 0.070 0.736 0.080 0.702
Living with natural mother age 16
No 0.000 0.000
Yes –0.440 0.013 –0.433 0.015
Father interested in child s education age 16
No 0.000 0.000
Yes –0.326 0.008 –0.332 0.008
Unknown –0.279 0.018 –0.270 0.023
School abilities at age 16
Half or more of subjects above CSE 1/O levels 0.000 0.000
More than half of subjects below CSE 1/O levels 0.375 0.023 0.339 0.041
Unknown 0.425 0.007 0.389 0.014
Source: Authors’ analysis based on survey data.
behavioural problems at age 16 had a significantly higher risk of poor mental health
at age 33. To summarise, most control variables, both at age 33 and in childhood,
are associated with mental health in the hypothesised directions.
Family Type
Of particular interest were the effects of family type. Figure 8.1 graphs the regres-
sion coefficients for the five family types (as in Model 1, Table 8.5). Compared to
those in first families, the risk of poor mental health is significantly higher for lone
parents and all three types of stepfamily. For lone parents, the odds of poor mental
health are twice as high (exp (0.692) = 2.0) as for those in first families, and for
dual stepfamilies 2.4 times as high (exp (0.869) = 2.38). The difference between
people in first families and people in families with only one stepparent is smaller,
yet significant at the 95% confidence interval level. These initial results are consis-
tent with the hypothesis that the various strains involved in stepparenting may result
in poorer mental health for both stepparents and their partners compared to those in
first families, and that these strains are even stronger in more complex stepfamilies
where both partners have stepchildren.
However, as we argued earlier, it is also possible that those prone to poorer men-
tal health may be selected into stepfamilies, and we showed in section “Research
Questions and Data” that those with a high HBS score at age 16 were more likely
to end up in a stepfamily at age 33 than those with a low HBS score at age 16.
Therefore, we controlled for behavioural problems in adolescence in Model 2, by
1.4
95% comparison interval
1.2
Estimated logistic effect
(regression coefficient)
0.8
0.6
0.4
0.2
–0.2
First Lone Stepparent Partner Dual
family parent of stepfamily
stepparent
Fig. 8.1 Poor mental health at age 33 (1991) by family status (derived from Model 1,
Table 8.5; Source: Authors’ analysis based on survey data.)
including an interaction between family type at age 33 and behavioural problem

status at age 16.
Figure 8.2 shows the coefficients calculated from the main effects and inter-
action effects of family type (at age 33) and behavioural problem status (at age
16) from Model 2 (Table 8.4). Of those who had few behavioural problems at age
2
HBS<7 HBS>=7 95% comparison interval
Estimated logistic effect
(regression coefficient)
1.5
0.5
–0.5
First Lone Stepparent Partner Dual
family parent of stepfamily
stepparent
Fig. 8.2 Poor mental health at age 33 (1991) by family status (derived from Model 2,
Table 8.4; Source: Authors’ analysis based on survey data.)
16 (HBS< 7), only lone parents and those couples where both partners were step-
parents had significantly worse health than those in first families at age 33. This
fits in with the finding from earlier studies that stepparent-stepchild relationships
were more distant if both parents had their own children living in the household
(Coleman and Ganong, 1990). In the more common stepfamilies, where only one
of the adults was a stepparent, neither the stepparents nor their partners had signifi-
cantly worse health than those in first families, if they had few behavioural problems
at age 16.
On the other hand, the risk for those who had a high HBS score at age 16 was
considerably higher in every family status category. Comparing across the categories
for those with high HBS scores at age 16, we find that lone parents, stepparents and
partners of stepparents and partners who were both stepparents, all have a signif-
icantly higher risk of poor mental health at age 33 than their counterparts in first
families (although only significant at the 90% confidence interval level for partners
of stepparents and dual stepfamilies; note that the larger confidence interval for dual
stepfamilies will partly be caused by the small number of observations in this cate-
gory (n = 55)). Hence our results suggest that those who had behavioural problems
in adolescence are much more likely to suffer poor mental health in later life, but
that those who end up living in stepfamilies (and lone parent families) suffer a sig-
nificantly heavier burden on their mental health.
Finally, we compare those with and without behavioural problems at age 16
within each family type. For those in first families who had high HBS scores at
age 16 the risk of having poor mental health at age 33 was significantly higher than
for those in first families with low HBS scores at age 16. Also, lone parents, step-
parents and partners of stepparents have a significantly higher risk of having poor
mental health if they had high HBS scores at age 16 than if they had low scores.
For dual stepfamilies there is no difference in poor mental health risk for those with
and without behavioural problems in adolescence. For both groups the risk is high,
indicating that living in a stepfamily with children from both partners’ previous rela-
tionships leads to stress and worries, no matter what the prior mental health status.
The next five hypotheses, concerning the characteristics of the adults and the
children in stepfamilies, were tested in multivariate models with the same set of
control variables as shown in Table 8.4. The effects of the control variables in all
models were very comparable to those in Table 8.4 and are not discussed further.
Characteristics of the Adults in Stepfamilies
We hypothesised that being poor affects the mental health of those in stepfami-
lies, because financial problems are a primary source of stress. We measured socio-
economic background in three ways: highest completed level of education, social
class and economic activity status (income information in the NCDS is not suitable
for this present analysis as it includes a lot of missing data, so we did not use it).
Each dimension shows the same picture, but the results for labour market activity
are the most marked, and remain largely significant after controlling for background
variables. The results show that people who do not work have a higher risk of poor
mental health than people who work. Among those who work, those working part-
time have a slightly elevated risk of poor mental health compared to full-time work-
ers, but only in the model without control variables. Overall, our findings confirm
that adults in stepfamilies with poorer socio-economic circumstances have a higher
risk of having poor mental health. However, we also find this effect for adults in first
families and lone parent families, so it is not exclusively true for stepfamilies. When
we control for background variables, the effects of level of education and social
class remain in the same direction, but their magnitude decreases and they become
largely insignificant.
Our next hypothesis was that the younger the stepparent, the greater the nega-
tive impact of stepparenting on mental health. In the NCDS, it is not possible to
test hypotheses about age differentiation directly, because NCDS cohort members
are all the same age. However, we explored this by comparing results at age 33
(NCDS sweep 1991) and age 41/42 (NCDS sweep 2000). Comparison of the results
shows that differences in mental health between those in first families and those in
stepfamilies were larger at age 33 than at age 41/42, confirming our hypothesis. Of
particular interest is the finding that at age 33, adults in stepfamilies have a higher
risk of poor mental health than those in first families, and this difference has disap-
peared at age 41/42 (although, only for those with behavioural problems at age 16
in the model with control variables). When we compare those who did and did not
have behavioural problems in adolescence within each family type, we also see that
differences were bigger at age 33 than at age 41/42.
Last, we find some confirmation for our hypothesised effect of traditional atti-
tudes on the likelihood of poor mental health for adults in stepfamilies. We did
this analysis on the data at age 41/42, with attitudes as measured at age 33 (prior
attitudes should provide a better explanation of current behaviour than current atti-
tudes, as the direction of effect is difficult to determine if both are measured con-
currently). A traditional attitude was measured as a cumulative score on several
attitudinal items such as “wives who don t have to work should not do so”. The
results show that adults in lone parent families and dual stepfamilies who held more
traditional views at age 33 have a slightly higher risk of poor mental health at age
41/42 than their counterparts who held less traditional views (although, only signif-
icant for those who had no behavioural problems in adolescence). For people in first
families and ‘single’ stepfamilies, attitude does not affect mental health. For people
who had behavioural problems in adolescence, attitude has no effect for stepfami-
lies and lone parent families, but it does for first families (the more traditional, the
higher the risk of poor mental health).
Characteristics of the Children in Stepfamilies

The presence of a child born to both parents in a stepfamily was expected to reduce
the risk of poor mental health, because it can cement the bond between family mem-
bers and provide more role clarity to the stepparent. We found marginal support for
this, as having common children did reduce rates of mental illness, but none of the
differences between adults in stepfamilies with and without common children were
statistically significant, even in the models without control variables.
The presence of adolescent children (age 13–17) in the household does increase
the likelihood of poor mental health, but this is true for all family types, not just
stepfamilies. In fact, the effect is greatest for adults in first families. When we con-
trol for background variables, the differences between families with and without
adolescent children disappears, except for those in first families and those in dual
stepfamilies who had no behavioural problems in adolescence.
Lastly, we studied the effect of having non-resident children (that is, either or
both adults in the stepfamily have a child or children who live(s) somewhere else,
usually with the other biological parent) on mental health. We expected that hav-
ing non-resident children would increase the risk of mental health problems. Firstly,
for the parent him/herself because (s)he is always in the shadow of the parent with
whom the children live, and because (s)he may miss the child(ren) and suffer from
feelings of guilt that (s)he does not spend more time with them. The stepparents of
non-resident children may also find it difficult to cope with stepchildren who they
only see infrequently, and to support their partner. Our results show a consistent
effect of having non-resident children in both the models with and without control
variables. Lone parents and adults in stepfamilies with non-resident children are
more likely to have mental health problems than their counterparts who do not have
non-resident children. The differences between those with and without non-resident
children are particularly large among people who had behavioural problems in
adolescence.
Summary and Conclusions
This study has shown that there are significant differences in mental health between
adults in first families and different types of stepfamilies. Stepparents themselves,
and their partners, are more at risk of having poor mental health than their coun-
terparts in first families. When both adults in a stepfamily are stepparents to each
other’s children, their mental health is even more at risk.
However, as we also showed, part of this association is due to the fact that adults
who live in stepfamilies are more prone to poor mental health in the first place.
This ‘proneness to poor mental health’ was captured using a variable that measured
behavioural problems at age 16, which is a good predictor of mental health problems
in later life. Those who had behavioural problems at age 16 were more likely to end
up in stepfamilies, and therefore make up a larger proportion of the stepfamily pop-
ulation than they do of the first family population. The temporal ordering of these
events means that cross-sectional data would not allow this issue to be explored.
Evenson and Simon (2005), for example, show a higher likelihood of depression
among some types of stepparents compared to parents in first families, but they
acknowledge in their discussion that this may partly be due to selection into and
out of stepfamilies of people with different prior mental health status (interestingly,
they speculate that it may be people with better initial mental health who select
themselves into stepfamilies).
Once we took prior mental health into account, we found that stepfamily life is
particularly harmful to the mental health of people who had behavioural problems in
adolescence. They appear to find it hard to cope with the role of being a stepparent,
or the partner of a stepparent, possibly because of the problems and tensions in step-
families that are largely absent in first families. Adults who did not have behavioural
problems as an adolescent only have an increased risk of poor mental health if they
live in dual stepfamilies or lone parent families.
Several other elements of stepfamilies were also shown to account for the adverse
effect of stepfamily life on mental health. When stepfamilies have non-resident chil-
dren, that is, one (or both) of the adults has children who live in another house-
hold, this increases the risk of poor mental health. The age of the stepparent also
mattered: the younger they were, the higher the likelihood of poor mental health.
Other factors increasing the likelihood of poor mental health in stepfamilies are: not
working, or working part-time (compared to working full-time), and having adoles-
cent (step)children in the household. However, these factors also increase the risk of
poor mental health for adults in the other family types (first families and lone parent
families), and are thus not only true for stepfamilies.
This study is one of the first of its type to compare the risk of mental health prob-
lems for stepparents, or partners of stepparents, compared to those in first families.
Although we could not directly compare our outcomes with any previous studies,
the results do conform to studies on related topics. Studies arguing that stepfamilies
are more stressful environments than first families (such as Brown and Booth, 1996;
Pryor and Rodgers, 2001) seem to be supported by our findings. Also, the worse
mental health of people in stepfamilies may affect marital quality and marital stabil-
ity of couples in stepfamilies. This would provide an additional explanation for the
markedly higher divorce rates found in remarriages that involve stepchildren. Such
marriages have been found to suffer from worse marital quality (e.g. Coleman and
Ganong, 1990; Pasley et al., 1993), but the explicit link with mental health has not
been made in such studies.
It is also possible that our results underestimate the adverse effect of stepfam-
ily life on mental health. If those who find it hard to cope with being a (partner
of a) stepparent are more likely to end the relationship, they will be ignored in
our analysis. While their mental health may have been most seriously affected by
step-parenthood, leaving the relationship may have caused improvements making it
difficult to estimate the effect of stepparenting. Only long-running panel data (with
frequent waves) would allow us to examine such a case, as we would be able to
observe people entering and exiting different types of households over their life
course, and relate their (changing) family type to their mental health status.
Our results demonstrate the value of longitudinal analysis, which has allowed the
subtlety of the relationship between stepparenting and mental health to be explored
and we support Coleman et al.’s (2000) plea for more longitudinal analyses of com-
plex family circumstances. Clearly, the health outcomes for stepparents and their
partners would appear to deserve more attention and the results of such analyses
may be of interest to family researchers, as well as clinicians and counsellors work-

ing with adults who are struggling with stepparenting issues.
References
Allan, G. (1999) The Sociology of the Family, Oxford, Blackwell.
Amato, P.R. (2000) The consequences of divorce for adults and children, Journal of Marriage and
the Family, 62: 1269–1287.
Ambert, A.M. (1986) Being a stepparent: Live-in and visiting stepchildren, Journal of Marriage
and the Family, 48: 795–804.
Avlund, K., Damsgaard, M.T. and Holstein, B.E. (1998) Social relations and mortality. An eleven
year follow-up study of 70-year-old men and women in Denmark, Social Science and Medicine,
47: 635–643.
Berkman, L.F. and Syme, S.L. (1979) Social networks, host resistance, and mortality: A nine-
year follow-up study of Alameda County residents, American Journal of Epidemiology, 109:
186–204.
Berman, C. (1980) Making it as a Stepparent: New Roles/New Rules. Garden City (NY),
Doubleday.
Bird, C.E. and Fremont, A.M. (1991) Gender, time use, and health, Journal of Health and Social
Behavior, 32: 114–129.
Brown, S.L. and Booth, A. (1996) Cohabitation versus marriage: A comparison of relationship
quality, Journal of Marriage and the Family, 58: 668–678.
Chase-Landsdale, P.L., Cherlin, A.J. and Kiernan, K.E. (1995) The long-term effects of parental
divorce on the mental health of young adults: A developmental perspective, Child Development,
66: 1614–1634.
Cherlin. A.J. (1978) Remarriage as an incomplete institution, American Journal of Sociology, 86:
634–650.
Cherlin, A.J., Chase-Lansdale, P.L. and McRae, C. (1998) Effects of parental divorce on mental
health throughout the life course, American Sociological Review, 63(2): 239–249.
Clingempeel, G. (1981) Quasi-kin relationships and marital quality, Journal of Personality and
Social Psychology, 41: 890–901.
Coleman, M. and Ganong, L.H. (1990) Remarriage and stepfamily research in the 1980s: Increased
interest in an old family form, Journal of Marriage and the Family, 52: 925–940.
Coleman, M., Ganong, L. and Fine, M. (2000) Reinvestigating remarriage: Another decade of
progress, Journal of Marriage and the Family, 62: 1288–1307.
Del Carmen, M., Alcon, G., Pedersen, J.M., Maria, A. and Gonzalez, C. (2002) Greenlandic family
structure and communication with parents: Influence on schoolchildren’s drinking behaviour,
International Journal of Circumpolar Health, 61: 319–331.
Duncan, G.J. and Hoffman, S.D. (1985) A reconsideration of the economic consequences of marital
dissolution, Demography, 22: 485–497.
Ermisch, J. and Francesconi, M. (2000) The increasing complexity of family relationships: Life-
time experience of lone motherhood and stepfamilies in Great Britain, European Journal of
Population, 16: 235–249.
Evenson, R.J. and Simon, R.W. (2005) Clarifying the relationship between parenthood and depres-
sion, Journal of Health and Social Behaviour, 46: 341–358.
Fergusson, D.M., Lynskey, M.T. and Horwood, L.J. (1994) The effects of parental separation, the
timing of separation and gender on children’s performance on cognitive tests, Journal of Child
Psychology and Psychiatry and Allied Disciplines, 35: 1077–1092.
Ferri, E. and Smith, K. (1998) Stepparenting in the 1990s, Family Policy Studies Centre, London.
Finch, N. (2002) Demographic Trends in the UK, Social Policy Research Unit, University of York,
York.
Firth, D. (2003) Overcoming the reference category problem in the presentation of statistical mod-
els, Sociological Methodology, 33(1): 1–18.
Flouri, E. and Buchanan, A. (2002) The role of father involvement in children’s later mental health,
Journal of Adolescence, 26: 63–78.
Gardner, J. and Oswald, A. (2004) How is mortality affected by money, marriage, and stress?
Journal of Health Economics, 23: 1181–1207.
Gayle, V. and Lambert, P. (2007) Using quasi-variance to communicate sociological results from
statistical models, Sociology, 41: 1191–1207.
Haskey, J. (1994) Stepfamilies and stepchildren in Great Britain, Population Trends, 76: 17–28.
Hetherington, E.M., Arnett, J.D. and Hollier, E.A. (1988) Adjustment of parents and children in
remarriage, in Wolchick, S.A. and Karoly, P. (eds.) Children of Divorce, University of Chicago
Press, Chicago, pp. 67–107.
Hetherington, E.M. (1991) Presidential address: Families, lies and videotape, Journal of Research
on Adolescence, 1: 323–348.
Hetherington, E.M. and Clingempeel, W.G. (1992) Coping with marital transitions: A family sys-
tems perspective, Monographs of the Society for Research in Child Development, 57: 2–3 (serial
no. 227).
Hetherington, E.M. and Jodl, K.M. (1994) Stepfamilies as settings for child development, in Booth,
A. and Dunn, J. (eds.) Stepfamilies: Who Benefits? Who Does Not?, Lawrence Erlbaum Asso-
ciates, Hillsdale (New Jersey), pp. 55–80.
Hirst, M.A. and Bradshaw, J.R. (1983) Evaluating the malaise inventory: A comparison of mea-
sures of stress, Journal of Psychosomatic Research, 27(3): 193–199.
Hope, S., Power, C. and Rodgers, B. (1999) Does financial hardship account for elevated psycho-
logical distress in lone mothers?, Social Science and Medicine, 49: 1637–1649.
Joshi, H., Cooksey, E., Wiggins, R., McCulloch, A., Verropoulou, G. and Clarke, L. (1999) Diverse
family living situations and child development: A multi-level analysis comparing longitudinal
evidence from Britain and the United States, International Journal of Law, Policy and the Fam-
ily, 13: 292–314.
Keirnan, K. and Mueller, G. (1998) Who are the divorced, and who divorces? CASEbrief 6, Centre
for Analysis of Social Exclusion, London.
Mekos, D., Hetherington, E.M. and Reiss, D. (1996) Sibling differences in problem behaviour and
parental treatment non-divorced and remarried families, Child Development, 67: 2148–2165.
Neff, J.A. and Schluter, T.D. (1993) Marital status and depressive symptoms: The role of
race/ethnicity and sex, Journal of Divorce and Remarriage, 20: 137–160.
Pagani, L., Tremblay, R.E., Vitaro, F., Kerr, M. and McDuff, P. (1998) The impact of family transi-
tion on the development of delinquency in adolescent boys: A 9-year longitudinal study, Jour-
nal of Child Psychology and Psychiatry and Allied Disciplines, 39: 489–499.
Palisi, B.J., Orleans, M., Caddell, D. and Korn, B. (1991) Adjustment to stepfatherhood: The effects
of marital history and relations with children, Journal of Divorce and Remarriage, 14(3/4):
89–106.
Pasley, K., Dollahite, D.C. and Ihinger-Tallman, M. (1993) Clinical applications of research find-
ings on the spouse and stepparent roles in remarriage, Family Relations, 42: 315–322.
Pong, S.L. (1997) Family structure, school context, and eighth grade math and reading achieve-
ment, Journal of Marriage and the Family, 59: 734–746.
Power, C., Manor, O., Fox, J. and Fogelman, K. (1988) Health selection: An explanation of
social inequalities in health in young adults? NCDS Working Paper 28, City University,
London.
Pryor, J. and Rodgers, B. (2001) Children in Changing Families: Life After Parental Separation,
Blackwell, London.
Ram, B.L. and Hou, F. (2003) Changes in family structure and child outcomes: Roles of economic
and familial resources, Policy Studies Journal, 31: 309–330.
Richards, M., Hardy, R., and Wadsworth, M. (1997) The effects of divorce and separation on
mental health in a national UK birth cohort, Psychological Medicine, 27: 1121–1128.
Rutter, M., Tizard, J. and Whitmore, K. (1970) Education, Health and Behaviour, Longman,
London.
Seeman, T.E., Kaplan, G.A., Knudsen, L., Cohen, R. and Guralnik, J. (1987) Social network ties
and mortality among the elderly in the Alameda County Study, American Journal of Epidemi-
ology, 126: 714–723.
Shouls, S., Whitehead, M., Burstrom, B. and Diderichsen, F. (1999) The health and socio-economic
circumstances of British lone mothers over the last two decades, Population Trends, 95: 41–46.
Simpson, B. (1994). Bringing the ‘unclear’ family into focus: Divorce and re-marriage in contem-
porary Britain, Man, New Series, 29: 831–851.
Smart, C. (2003) Introduction: New perspectives on childhood and divorce, Childhood, 10:
123–129.
Teachman, J.D., Paasch, K. and Carver, K. (1996) Social capital and dropping out of school early,
Journal of Marriage and the Family, 58: 773–783.
Trovato, F. and Lauris, G. (1989) Marital status and mortality in Canada: 1951–1981, Journal of
Marriage and the Family, 51: 907–922.
University of London (2000). Institute of Education. Centre for Longitudinal Studies, National
Child Development Study: Sweeps 1 (1965), 2 (1969) and 3 (1974) [computer file]. National
Birthday Trust Fund, National Children s Bureau and City University Social Statistics Research
Unit [original data producer(s)]. UK Data Archive, University of Essex, Colchester [distribu-
tor], 25 July 2000 (2nd Edition). SN: 3148.
University of London (2007). Institute of Education. Centre for Longitudinal Studies, National
Child Development Study: Sweep 5, 1991 [computer file]. City University. Social Statistics
Research Unit, [original data producer(s)]. Colchester, Essex: UK Data Archive [distributor],
January 2007. SN: 5567.
Utting, D. (1995) Family and Parenthood: Supporting Families, Preventing Breakdown, Joseph
Rowntree Foundation, York.
Wade, T.J. and Pevalin, D.J. (2004) Marital transitions and mental health, Journal of Health and
Social Behavior, 45: 155–170.
Wadsworth, J., Burnell, I., Taylor, B. and Butler, N. (1985) The influence of family type on chil-
dren’s behaviour and development at five years, Journal of Child Psychology and Psychiatry,
26: 245–254.
Walper, S. (1995) Family relationships and adolescents social development in nuclear, single
mother, and stepfather families, Zeitschrift für Entwicklungspsychologie und Pädagogische
Psychologie, 27: 93–121.
White, L.K. and Booth, A. (1985) The quality and stability of remarriages: The role of stepchildren,
American Sociological Review, 50: 689–698.
Zill, N. (1988) Behavior, achievement, and health problems among children in stepfamilies: Find-
ings from a national survey of children health, in Hetherington, E.M. and Arasteh, J. (eds.)
The Impact of Divorce, Single Parenting and Stepparenting on Children, Lawrence Erlbaum
Associates, Hillsdale, New Jersey, pp. 325–368.
Chapter 9
Grandparents and the Care
of Their Grandchildren
Alison Smith Koslowski
Introduction
Across Europe, certain grandparents are being relied upon to provide informal care
for their grandchildren, particularly as part of a package of early-years childcare
provision (Gray, 2005; Lewis et al., 2008). The increased participation of women
in the labour market (Crouch, 1999), without concomitant and equivalent increased
participation of men in informal care (Hook, 2006) means that the demand for infor-
mal care is likely to increase whilst supply decreases. Evidence suggests that grand-
parents are the informal carers of choice after the parents themselves (Wheelock
and Jones, 2002). However, certain interest groups representing grandparents have
raised the concern that grandparents may be making a rather larger contribution to
the care and upbringing of their grandchildren than they had anticipated (e.g. Age
Concern, 2006). Regular primary care of small children is a time consuming activity,
which is both physically and emotionally demanding and as such, not a task that
grandparents had necessarily envisaged as a core part of their grandparental role.
There has been an increasing focus on informal care provision by mothers and
their subsequent labour market and other outcomes (e.g., Joshi et al., 1999) and less,
but some work, on informal care provision by fathers (Smith Koslowski, 2008).
There has also been recognition of the prevalence of the informal care provided
by older people for other older people, whether care for a spouse or for very
elderly parents (e.g., Bolin et al., 2007). However, there has been little research
in Europe into informal care provision by grandparents for grandchildren. Esti-
mates for the value of unpaid childcare provided by grandparents in the UK suggest
considerable activity, ranging from £3,886 million a year (Age Concern, 2004),
to £50 billion (Rowlatt, 2007) to £220 billion (Greengoss, 2007). Such macro-
economic accounting relies in part on information about the time use of older
persons.
A.S. Koslowski (B)

School of Social and Political Science, University of Edinburgh, Edinburgh, UK
e-mail: Alison.J.Smith@ed.ac.uk

172 A.S. Koslowski
For grandparents, there may well be financial implications associated with car-
ing informally for their grandchildren, although the costs and benefits will certainly
extend beyond the financial, including implications for health, volunteer work and
leisure (for example, Bamford et al., 1998). For grandchildren, distinct develop-
mental outcomes for children are associated with being cared for by a grandparent
(e.g., Hansen and Hawkes, 2009). In order to calculate and thus understand the
costs and benefits of grandparental informal childcare, the first step is to obtain
data, which tell us about the amount of time grandparents spend looking after their
grandchildren.
If policy makers are concerned to increase the supply of informal care, the impor-
tance of support, financial and otherwise for individual care providers, requires con-
sideration. It may be that an increased reliance on grandparents as informal carers is
in contradiction with other policy goals associated with ‘active ageing’, concerned
with raising the employment rates of older people (Lewis et al., 2008). However, a
broader definition of ‘active ageing’ may incorporate and recognise such informal
care work. The provision of informal grandparental childcare is likely to be gen-
dered. Whilst grandfathers as well as grandmothers are shown to be active child
carers, grand-maternal care is more prevalent (Gray, 2005). Such gendered informal
care is likely to be one of several contributory factors leading to women in later life
experiencing an increased risk of poverty than men (Price, 2006). A review of the
time spent looking after children by grandparents, is a first step in the process of
assessing the implications of grandparental childcare for the grandparents involved,
be they financial and otherwise, positive or negative.
The role of grandparents and their contribution to society is increasing in pro-
file, not least due to there being more grandparents than ever before as mortality
decreases, despite a substantial decline in fertility (cf. Murphy and Grundy, 2003),
and due to the increase in divorce rates and subsequent extended family, result-
ing in more complex grandparent and step-grandparent relationships. Indeed, an
array of dynamic demographic factors including increasing life expectancy, delayed
childbearing, decreasing fertility rates, teenage pregnancy rates and increased inter-
generational geographic mobility contribute to the complexity of this issue. There
has been somewhat of a lull in the latter part of the twentieth century as regards
research into extended family relations (Dench and Ogg, 2002). As such, it is diffi-
cult to contextualise grandparental time in recent historical perspective and to know
whether there is a revival of extended family relations, or if ties are as active as they
ever were.
This chapter presents a cross-national review of the time that grandparents report
spending on childcare in Western Europe. Estimates are largely drawn from data
representing the time period 1994–2005. First, the various forms of grandparental
childcare are presented, introducing the focus of the chapter, namely whether the
grandparent is acting as an informal childcare provider typically whilst the parents
go out to work, although grandparents might also provide informal care as respite
for non-working parents. Second, the extent to which informal grandparental child-
care can be expected to vary across Western Europe is discussed. Inevitably, dif-
ferent welfare state systems create varying circumstances with regard to childcare
9 Grandparents and the Care of Their Grandchildren 173
provision and parental household employment patterns. A discussion of the com-

plexity of the measurement of grandparental childcare time then precedes a review
of currently available estimates.
Grandparents as Informal Childcare Providers
The initial picture of grandparents spending time with their grandchildren per-
haps conjures up thoughts of Sunday afternoon family visits and proudly displayed
school photographs and no doubt, this is a grandparental model for many families.
A concern of this chapter however, is to highlight a particular category of grand-
parental childcare; that where the grandparent is a primary informal carer on a
regular basis. The use of the term ‘primary’ indicates the parents’ absence typ-
ically for a period of hours at any one time, but in some cases more long term
to the extent of surrogate parenting. Grandparental childcare that extends beyond
multi-generational family visits and holidays can be said to exist in two forms. The
first form is where the grandparent becomes a co-residential surrogate parent to the
grandchild. The second form is where the grandparent becomes the primary infor-
mal childcare provider. The latter is the more commonplace. Both across and within
these categories of grandparental childcare, grandparents are a heterogeneous group,
not least in terms of health, wealth and age.
Co-residential childcare carried out by grandparents acting as surrogate parents
is the form of grandparental childcare that has attracted most research attention
in recent years, though much of this literature is non-European (Hirshorn, 1998;
Richards, 2001; Grandparents plus and Adfam, 2006; Brennan and Cass, 2007;
MacKenzie, 2007). This particular form of grandparental childcare occurs in the
event of parental failure, whether as the result of the death of the parents, or more
commonly, a parental custodial prison sentence or substance abuse. In such cases,
grandparents will frequently step in and take on the role of surrogate parent. Social
workers may refer to grandparental care as a subset of kinship care. The support
of grandparents has been identified as a significant protective factor for children
(Kroll, 2007) as kinship placements are more likely to enable children to remain
within a family network than placements with non-kin carers (Broad, 2004). How-
ever, complications for the grandparents may be caused by uncertain legal status,
prior relationship difficulties with the parents, and factors such as their health,
wealth and age (as described in Kroll, 2007). This group of grandparents are perhaps
particularly vulnerable and in need of state support. The number of grandparents act-
ing as surrogate parents is relatively small; Clarke and Cairns (2001) estimate that
1% of grandparents in Britain had grandchildren living with them on such a basis.
A small number of grandparents are co-resident with their children and grand-
children. The percentage of parental households with resident grandparents and
dependent children ranges from just 0.12% in the Netherlands to 15.6% in Portugal
(Table 9.1). Single grandparents living with children are more likely to look after
grandchildren, compared to those who do not live with their children (Croda and
Gonzalez-Chapela, 2005).
174 A.S. Koslowski
Table 9.1 Co-resident parental households containing grandparents and dependent children
% households
with
grandparents N
Portugal 15.64 25,566

Greece 14.98 24,190
Austriab 14.13 13,066
Spain 11 31,584
Luxembourga 6.85 1,984
Italy 6.41 36,204
Ireland 5.51 17,146
UKa 2.22 6,852
Germanya 1.89 7,288
Belgium 1.5 14,036
Finlandc 1.45 14,040
France 1.29 29,234
Denmark 0.47 10,952
Netherlands 0.12 22,124
Source: ECHP pooled sample 1994–2001 inclusive as available,
author’s calculations.
a
Data coverage 1994–1996 inclusive.
b
c
More commonplace, but less evident in the literature are those non-residential
grandparents providing regular informal childcare in the absence of the parents.
Informal childcare may be met with some cash payment, but frequently other forms
of reciprocity and redistribution come into play, commonly fulfilled over a long
period of time. In contrast, formal childcare can be defined as that provided by
external private service providers, the state or by not-for-profit organisations (Land,
2002). In the latter quarter of the last century and into the present decade, there
has been significant growth in the employment rates of mothers (Crouch, 1999).
There is cross-national variation, but this is an overall European trend (Lewis
et al., 2008). The Lisbon Agenda, as set out by the European Council in March
2000, is an action and development plan for the European Union which, among
other priorities, promotes the active employment of women with dependent chil-
dren, including lone mothers. Many of these women will have young children who
require childcare, and grandparents are an important source, as recognised by the
UK Government in its 1998 consultation paper Supporting Families. Indeed, much
evidence points to grandparents being the most important source of such care in
some countries, including the UK (Douglas and Ferguson, 2003; Bradshaw et al.,
2008).
There is as yet little to no financial assistance for grandparents who are caring
for their grandchildren as primary informal care providers (Douglas and Ferguson,
2003; Rowlatt, 2007). Currently, such grandparent childcarers are rarely eligible for
statutory support. In Hungary, grandparents have the possibility to take over some
of the paid parental leave afforded to parents (Szikra, 2008). Portugal offers the
opportunity for the mother of an under-age teenage mother to take parental leave in
her place (OECD, 2004).
Informal care is for many parents a positive choice (Land, 2002). Informal care
arrangements may be better able to withstand the complex timetabling associated
with atypical working hours (Le Bihan and Martin, 2004). This preference may be
associated with the affordability, availability and quality of formal childcare pro-
vision, all characteristics which vary cross-nationally. Whilst many grandparents
provide regular primary informal childcare willingly, there is some evidence to sug-
gest that a significant minority would prefer not to take on such an intense role, but
rather keep to a less involved grandparental status (Dench and Ogg, 2002; Douglas
and Ferguson, 2003).
There is considerable heterogeneity within the members of the population who
might describe themselves as grandparents. It is difficult to generalise the life stage
at which grandparents may find themselves, as grandparenting is a contingent and
complex process (Kemp, 2003). Becoming a grandparent can occur at anytime from
around 35 years to 75 years of age and beyond. Evidence suggests that 27% of all
grandparents across Western Europe are below the age of 60, and so not of retire-
ment age (Hank and Buber, 2007). The high prevalence of teenage pregnancy in
the UK, has led to a relatively high prevalence of young grandparents, themselves
still of child bearing age (BBC3, 2008). This group of grandparents are particularly
likely to be turned to for informal primary childcare (Anderson et al., 2007). At
the other end of the socio-economic spectrum, delayed childbearing leads to a rel-
atively high prevalence of older grandparents. The opportunity costs, in particular
the financial implications of becoming a grandparent at various stages along the life
course will differ according to whether a person is still active in the labour market,
non-employed or retired.
Grandparental childcare is a gendered activity. Various studies show that
maternal grandparents appear to be more significant in the life of grandchildren
than paternal grandparents, and that grandmothers appear to be more involved
than grandfathers (Dench and Ogg, 2002; Douglas and Ferguson 2003). Adding
further complexity, many grandchildren will have more than four grandparental
figures in their lives (e.g. Kemp, 2003; Bradshaw et al., 2008). Largely this is the
result of parental divorce and re-partnering as well as grandparental divorce and
re-partnering, both of which lead to step-grandparenthood. Social grandparenting
refers to those who take on a grandparenting role without necessarily being bound
by biological ties. Depending on survey construction, studies often implicitly collect
data from social grandparents.
In brief, two forms of grandparental childcare have been outlined; surrogate par-
enting and primary informal childcare provision. The focus of this chapter is on the
latter category, predominantly because the former group is not large enough to be
picked up by most survey data collection due to relatively small sample sizes. Whilst
this does represent a particular form of grandparental contact and childcare, those
grandparents involved cannot be considered to be a homogenous group. The het-
erogeneity between those grandparents giving regular informal care has potentially
176 A.S. Koslowski
important consequences for the financial implications and other associated costs or
benefits to the provider of such care work.
Determinants of Cross-National Variation in Informal

Grandparental Childcare Provision
Cultural differences and nationally specific institutional structures will contribute
to cross-national variation in grandparental care. Furthermore, by comparing across
countries we can explore whether socio-economic factors are consistently corre-
lated with grandparental time spent on informal childcare provision. If this is found
to be the case, variation in prevalence of informal grandparental childcare provision
may be explained in part by compositional differences. In this section, the implica-
tions for grandparental care of variation in state intervention and parental household
employment patterns are particularly explored.
Grandparental care is likely to become increasingly important when parents, par-
ticularly mothers, are involved in employment. To the extent that children are a
public good, the state can ease the difficulties associated with combining parent-
hood and work. How it does so will affect the prevalence of grandparental childcare
as mediated by the demands on the parents. Government regulated childcare pro-
visions can be classified into three distinct categories (Ditch et al., 1996): firstly,
cash transfers (via fiscal policies, family allowances and other forms of social assis-
tance); secondly, provision of services (e.g., the public provision of childcare), and
thirdly, leave from the labour market. Generous cash transfers and provision of ser-
vices in particular will enable parents to draw on formal care rather than rely on
informal care by grandparents. If parents are better supported financially and in the
absence of very high childcare costs, working long hours for more money may prove
unnecessary. Statutory leave provisions are positively associated with the time that
mothers and fathers spend looking after their children (Smith and Williams, 2007).
If parents are able to spend more time caring for their children, again the burden on
grandparents will be reduced.
Assuming the involvement that grandparents may continue to have, even in light
of the better case scenarios described above, the state can also introduce forms of
additional social assistance for grandparental carers, and/or a form of (paid) leave
from the labour market.
The European Union provides an ideal opportunity for comparative work. Whilst
the member states share a common institutional framework, there is also internal
diversity in terms of the welfare states, the governing institutions, traditions and
cultures, which allows for variation in the explanatory variables. Additionally, there
is also a practical rationale for focusing on the EU-15: the data situation is optimal
for these countries.
There has been some considerable debate on whether there has been a Euro-
pean welfare state convergence or divergence (see Castles, 2004), and there is a
burgeoning literature on the more recent emergence of the European social model
(e.g., Esping-Andersen et al., 2002). The basic premise of a European social model
is that the EU institutions harmonise social policy in Europe so as to provide social
protection in a distinct fashion from that found elsewhere in the developed world
(Castles, 2004).
The Social Chapter was adopted at the European Council in December 1989 by
all member states except the UK (EC, 1990) and included the statement that “mea-
sures should also be developed enabling men and women to reconcile their occupa-
tional and family obligations” (CEC, 1989, p. 5, cited in Neilson, 1998, p. 72). In
December 1991, at Maastricht, the member states (with the exception of the UK)
agreed to implement the Social Chapter by means of the Protocol on Social Policy
and the Agreement on Social Policy, which was annexed to the Treaty on European
Union (EC 1992; Neilson, 1998), thus writing mother- and father-friendly policy
concerns firmly into the agendas of the member states.
A council recommendation in 1992 (92/241/EEC) was passed, in which member
states are encouraged to take and/or progressively encourage initiatives to enable
women and men to reconcile their occupational, family and upbringing responsibil-
ities arising from the care of children (Article 1). A recommendation such as this is
an example of soft law, meaning that, unlike hard law, there are no sanctions on non-
compliance. No mention was made of grandparents being part of the solution. More
generally, the legal position of grandparents will vary across countries. Currently,
in English and Scottish law, grandparents have “no particular legal status underpin-
ning their relationship with their children” (Douglas and Ferguson, 2003, p. 42). In
contrast, Germany has specific rights of grandparental access, subject to the welfare
of the child, as does the USA (Douglas and Ferguson, 2003, p. 43).
Different welfare state systems inevitably lead to variation in the manner and
extent of provision for children. Gornick and Meyers (2003) identify how important
it is to develop an ‘end vision’ which honours the importance of earning and caring,
child well-being and gender equality. Crompton (1999) has conceptualised one such
society, a dual-earner-dual-carer society. As Gornick and Meyers note:
“This is a society that recognises the rights and obligations of women and men to engage
in both market and care work and one that values children s need for intensive care and
nurturance during their earliest years” (Gornick and Meyers, 2003, p. 12).
Europe is still far from such a dual-earner-dual-carer vision. Despite increas-

ing European harmonization considerable variation remains across countries with
regard to the extent of childcare provision, and consequently the reliance of parents
on informal grandparental care.
Across Europe, the emphasis has been on the labour market activation of moth-
ers and not on the care activation of fathers. In some countries, more help is given
however, to enable fathers to spend more time looking after their children via father-
friendly provisions (Smith and Williams, 2007). In the case that mothers are turning
to their parents for help, rather than to fathers, this can be said to be symptomatic
of a country holding less gender egalitarian attitudes. The suggestion is not that
fathers might become househusbands: that they might undergo a reversal of tra-
ditional roles. Rather, the suggestion is that fathers could be expected to provide
178 A.S. Koslowski
a similar level of care as provided by a mother, who was equally as active in the
labour market as the father. Clearly, this model is not plausible if both parents are
working long hours. As a normative idea, this model assumes the level of care a
parent could reasonably provide within the context of working a reduced or flexible
hours working week, with the partner doing the same, whilst they are having access
to affordable, high quality childcare.
State intervention notwithstanding, increased female labour market participation
makes new childcare strategies necessary (Wallace, 2002). However, patterns of
female labour market participation vary widely across Western Europe, particularly
in regard to hours worked (Daly, 2002). Part-time employment (under 30 hours per
week) is very important for mothers in the Netherlands, Germany, the UK, Ireland,
Austria and Belgium. In contrast, fewer than 10% of mothers of children under 12
work part time in Finland, Greece, Portugal and Spain (Lewis et al., 2008). Sum-
marizing broadly, Lewis et al. (2008) present three empirically observed national
parental household employment patterns. The first consists of those countries where
dual full-time working is the most usual arrangement and there are few single-earner
couples. Countries include: Denmark, Finland and Sweden. The second group con-
sists of those countries that polarize between dual full-time earning and single earner
families, with few part-time mothers. Countries in this group include Spain, Greece
and Portugal, although Portugal has higher rates of dual full-time earner couples.
The third group consists of countries where households with one full-time earner
and one part-time earner are prevalent, with other models also present. Countries in
this group include the Netherlands, Germany and the UK.
Finally, the extent to which parents and grandchildren have moved away from
their grandparents is likely to vary across countries. Geographical proximity is the-
oretically expected to increase the likelihood of informal grandparental childcare
provision and this is found to be empirically borne out (Hank and Buber, 2007;
Bradshaw et al., 2008). However, the perception of distance can also be culturally
specific. Living ‘nearby’ in France could mean the same distance as travelling to
the next region in the Netherlands. Furthermore, a certain number of grandparents
may also be commuting to care, although given the paucity of literature, this does
not appear to have been empirically explored, thus presenting an area for further
investigation.
Measurement of Grandparental Time
The parental role can be thought of as being comprised of two elements: caring and
providing. Resources devoted to children can then be understood as either financial
resources or as care resources. Grandparental roles can be thought of similarly. Inter-
generational financial resources can be measured in monetary terms through the use
of family expenditure and other household studies, but care resources are more dif-
ficult to take into account. One way to assess a grandparent’s involvement with care
is to consider the quantity of time that they spend looking after their grandchildren.
Grandparental time is defined as the amount of time spent by a grandparent looking
after their grandchild(ren). Grand-maternal time refers to grandmothers and grand-

paternal time to grandfathers.
Little is known about the impact of the time spent by (non-resident) grandparents
on children’s development, but some outcomes do appear to be distinct as com-
pared to children cared for by formal providers (Hansen and Hawkes, 2009). Grand-
parental time measurement, however, provides insight beyond child outcomes. It is
also a starting point for explaining socio-economic differences between the adults
involved. Whilst it may not say much about the quality of the childcare, it can nev-
ertheless be taken as an indicator of the intensity of involvement and thus inform
macro-economic accounting and individual-level opportunity cost calculations.
But how do studies go about measuring the allocation of time? One of the char-
acteristics of data on time allocation is that valid measurements are both difficult
and costly to obtain. Most time use data have to be based on self reported esti-
mates, which are prone to bias. There is always the possibility that a respondent
replies with what they consider to be an appropriate answer rather than with a true
account. Or, as Budig and Folbre (2004, p. 52) comment, “cultural norms as well as
social expectations affect the ways people perceive their own activities”. Research
has shown that husbands and wives will disagree in their perceptions of household
shared contributions to childcare and housework (Lee, 2005).
The most accurate way to record time use within a grandparental household
would be ludicrously intrusive; it would be to directly observe, perhaps using cam-
eras or a tagging device. Direct observation would also be likely to affect respon-
dent behaviour. As such, information on time use is typically collected either using
time sampling, with time diaries, or with retrospective survey questions (Gershuny,
2004). Time diaries are generally the method of choice for time use researchers,
administered to a sample of individuals in a population and organised in such a way
as to provide a probability sample of all types of days and of the different seasons
of the year. Time diaries are usually retrospective, asking respondents for a detailed
breakdown of the previous day, with responses coded according to a standard list
of activities (Juster and Stafford, 1991). More commonplace than time diaries are
retrospective survey questions about typical time use as part of a larger, more gen-
eral survey. As such, grandparents are directly asked how much time they spend
on certain activities such as grandparental childcare and housework. The validity
and reliability of data collected in time diary form is generally considered to be
superior (Juster and Stafford, 1991). However, there are many strong arguments
for using retrospective survey questions. Often, these questions are part of a nation-
ally representative, possibly longitudinal, annual household survey, rich in other
socio-economic information about the respondents. These studies are often well
resourced, particularly in terms of experienced research teams and are well doc-
umented. Estimates are generally reported in terms of hours per week or hours per
day, although occasionally hours per year are the unit of measurement.
Sometimes the boundary between what constitutes childcare and other forms of
unpaid work is blurred, for example, doing laundry and preparing family meals.
Even when it is the sole activity under consideration, childcare can be classified
as active, (also referred to as primary) or passive (also referred to as secondary).
180 A.S. Koslowski
Active childcare estimates capture only the time that grandparents report being
solely involved in caring for children. Passive childcare estimates include time
that the grandparent spends with children whilst simultaneously engaged in other
activities. Most data on grandparental care are not yet at such a stage as to offer a
differentiation between active and passive estimates.
There are currently two potential perspectives on grandparental time: time as
reported by the grandparents themselves, possibly in comparison with other per-
sons of a similar life stage but perhaps without grandchildren; and time as reported
from the perspective of the grandchildren, most probably mediated by the mid-
dle generation. Ideally, cross-generational data are required to fully understand the
intra-extended-family household dynamics of care.
Data from the perspective of grandparents can be collected as part of a gen-
eral household survey, but such surveys are likely to suffer from small samples of
grandparents, particularly those who are primary informal child carers. The Euro-
pean Community Household Panel (ECHP) which covers the period 1994–2001
for 15 countries (although not all countries for every survey year), provides some
information on the time respondents report spending looking after children on a
weekly basis. This variable has been used to consider how much time parents
spend looking after children (e.g. Smith and Williams, 2007). The ECHP provides
childcare information for co-resident grandparents but the numbers are negligibly
small in most countries (Table 9.1). Common to all household surveys however,
it does not pick up on family members living outside the household. This is not
the ideal vehicle then for exploring the majority of grandparental childcare activi-
ties. There are plans to address this in the new UK Household Longitudinal Study.
Time Use studies are another general population source of data on grandparental
care. The Multinational Time Use Study provides data from highly comparable
national surveys all drawing on a diary as the mode of data collection and has been
used to generate estimates for time spent looking after grandchildren (Gauthier and
Smeeding, 2003).
Other surveys do not aim to achieve coverage of the general population, but rather
take the older person (often defined as over 50 years) as their focus. This includes the
Survey of Health, Ageing and Retirement in Europe (SHARE), which collects har-
monised data for a range of countries, namely Austria, Belgium, Denmark, France,
Greece, Germany, Italy, the Netherlands, Sweden, Switzerland and Spain. Israel, the
Czech Republic, Poland, Slovenia and Ireland join in later waves of data collection.
There is also the English Longitudinal Study of Ageing (ELSA), which has many
questions in common with SHARE and also the U.S. Health and Retirement Study
(HRS). There is not currently a study of ageing for Scotland. A problem of col-
lecting time use information from older adults only for the purposes of grandpar-
ent research is that younger grandparents will be missed out. Cross-sectional data
may also be collected from the perspective of the older person. An example is
the British Social Attitudes 1998 module on the role of grandparents (Dench and
Ogg, 2002).
Birth cohort studies provide information on grandparental care, from the per-
spective of the child, as mediated by the parent. Currently, in the European arena,
such studies are limited in number. These are the Millennium Cohort Study and
Growing up in Scotland. A third cohort study, Growing up in France will also
soon be available (from 2010). The birth cohort studies follow a child from birth
or early childhood via the primary carer. In 98% of cases the primary carer is
defined as the mother, as a product of the survey design, which aims to col-
lect information about pregnancy and breastfeeding. The mother is asked about
proximity to and involvement of grandparents. Birth cohort studies from a medi-
cal perspective have been carried out in Denmark, Norway and the Netherlands.
There is also the Lifeways Cross-Generation Cohort Study in Ireland, but this is
on a much smaller scale than the British cohort studies and so cannot be con-
sidered as representative of the general population. All of the studies mentioned
above are longitudinal. The European Social Survey offers a cross-sectional cross-
national overview of grandparental childcare from the perspective of the mother in
2004/5.
The case where a grandparent steps in as a surrogate parent would appear to
be a small subset of grandparents, only 1–2% in the case of small children and
infants. Birth cohort studies would identify such a grandparent in this case as
‘primary caregiver’ and time spent by the grandparent might be better thought of
as a special form of parental time. Studies where the focus is on the older person are
less likely to distinguish between these two forms of care. As such, the data sources
described in the previous section are not the best source of information on this form
of grandparental care.
This discussion highlights that the measurement of parental time is a com-
plex activity, and that there are various approaches to procuring secondary data on
grandparental childcare. We now move to a discussion of the current estimates of
grandparental time.
Review of Estimates of Grandparental Time

Grandparents who provide informal care on a regular basis for their grandchildren
are of central interest in this review. However, general estimates, across the forms
of grandparental childcare, are also presented in so far as they are available. From
the grandchild’s perspective from the cohort studies, information on grandparental
time is collected in a variety of formats. A typical question is:
How often, on average, would the child actually see his/her grandparent(s)?
– Every day or almost every day
– At least once a week
– At least once a month
– At least once every three months
– Less often than that
– Never
(Adapted from Growing up in Scotland (GUS) main stage questionnaire, 2005).
182 A.S. Koslowski
More detailed information comes through the following set of questions on a

cohort study. Again this example is taken from Growing up in Scotland, but similar
questions are also asked in the Millennium Cohort Study.
Grandparents sometimes help out in practical ways with bringing up children. Can
you tell me how often, if at all, any of your child’s grandparents, help out in the
following ways?
. . . look after your child for an hour or more during the day
. . . babysit for your child during the evening
. . . have your child to stay overnight
. . . take your child on outings or daytrips without you (or your partner)
Again, the response options for this question would typically be:
– Every day or almost every day

– At least once a week
– At least once a month
– At least once every three months
– Less often than that
– Never
Using these questions as the source of information, reports from the Growing
up in Scotland and the Millennium Cohort Study find that the child’s grandpar-
ents are a key source of support. In Scotland, the vast majority (over 94%) of all
cohort children were in regular contact (at least once every three months) with all or
some of their grandparents (Bradshaw et al., 2008, p. 9). Almost half the children
in the GUS samples saw their grandparents daily or almost daily (Anderson et al.,
2007). Around four out of five (80%) parents in both the baby (aged approximately
10 months at the time of first interview) and toddler (aged approximately 34 months
at the time of first interview) cohorts in the study, said that the child’s grandpar-
ents looked after the child for an hour or more during the day at least occasion-
ally, including one-fifth where this was a daily or almost daily occurrence. Between
18–19% of cohort children were looked after by their grandparents during the
evening at least once a week, and 13% stayed with their grandparents overnight
at least once a week (Bradshaw et al., 2008). Twenty-nine per cent of lone parents
said that a child’s grandparent and/or grandparents looked after the child during the
day on a daily or almost daily basis compared with 17% of couple families.
More detailed hourly time use information comes through the following set of
questions in cohort studies. Again, this example is taken from GUS, but virtually
identical questions are also asked in the Millennium Cohort Study (MCS).
Do you currently get help with childcare for your child on a regular basis from any
of the providers or people listed on the card?
The child’s grandparent(s)

Another relative
Private crèche or nursery
Childminder
Local authority playgroup or pre-school
Local authority crèche or nursery
Community/Voluntary playgroup or pre-school
Private playgroup or pre-school
My ex-spouse or partner
The child(ren)’s older brother or sister
A friend or neighbour
Daily nanny who came to our house
Live-in nanny
Babysitter who came to our house
Workplace crèche or nursery
Family centre
Nursery class attached to primary school
Child-carer (provided via childcare agency)
Other childcare provider (PLEASE SPECIFY)
For roughly how many hours would your child be in their care in an
average week?:
The results from the GUS question above tell us that 34.5% of mothers were
depending on grandparents as their primary childcare provider when their child was
at 10 months (Table 9.2). The most common type of childcare provider used across
the samples was a grandparent. Around two-thirds of baby families and half of tod-
dler families using childcare report some arrangement with the child’s grandparents
(Anderson et al., 2007). The rate of use of grandparents for informal childcare is
highest in the most deprived neighbourhoods. In contrast, there is a significantly
lower use of nurseries, playgroups and childminders in these areas.
On average, those babies cared for by a grandparent were cared for 16 hours
per week (standard deviation 13 hours). This estimate is similar to research by Age
Concern that grandparents who provide childcare do so for an average of 15.9 hours
a week (Age Concern, 2004). This is very similar to reports from the MCS on child-
care use by working mothers when the children are aged 9 months, which estimates
that informal grandparent care (most frequently the maternal grandmother) accounts
for 35.3% of childcare (Hansen and Hawkes, 2009). Gray (2005) uses data from the
UK Time Use Survey 2000 to calculate weekly hours of grandparent childcare by
mother’s employment status and occupational group. This analysis was confined to
mothers of one child under 12 years old. Within occupational groups, differences by
employment status (not employed, part-time, full-time) again confirmed the greater
importance of grandparent childcare for lower-income mothers. The average weekly
hours of grandparent care averaged across all occupations for all mothers of one
child under 12 was 6.01 hours.
184 A.S. Koslowski
Table 9.2 Primary childcare provider at ten months in Scotland, 2005
Type of provider N Percent
Not using childcare 2,107 40.39

Grandparents 1,799 34.48
Private crèche or nursery 571 10.94
Childminder 279 5.35
Another relative 166 3.18
Local authority crèche or nursery 79 1.51
A friend or neighbour 64 1.22
My ex-spouse or partner 30 0.58
Local authority playgroup or pre-school 18 0.35
Daily nanny who came to our house 16 0.31
Workplace crèche or nursery 15 0.28
Private playgroup or pre-school 12 0.23
The child’s older brother or sister 10 0.20
Family centre 7 0.14
Community/voluntary playgroup 7 0.13
Babysitter who came to our house 4 0.07
Live-in nanny 4 0.07
Child-carer via agency 3 0.05
Nursery class attached to school 1 0.02
Other childcare provider 21 0.40
Total 5,217 (5,213 here 100 (99.9 due to
due to rounding rounding)
after weighting)
Source: GUS: Sweep 1 Birth Cohort (weighted); author’s calculations.
Similarly, on a cross-national scale, the European Social Survey 2004/5 asks the
question:
Thinking about the youngest child in the household, I would like to ask you about his/her
usual childcare, not counting lessons in school. By childcare I mean care carried out by
anyone other than yourself or your current husband/wife/partner.
Using this card, what is the main type of childcare that the youngest child receives?
The range of response options is similar to those outlined by the GUS ques-
tion example above. Again, one of the responses is ‘the child’s grandparent(s)’. The
responses from this survey are reported in Lewis et al. (2008) for 13 European coun-
tries. They find that between 6.2 to 61.7%, depending on the country, of mothers
in paid work, with a child aged 0–6 are having their main type of childcare
provided by grandparents (Fig. 9.1). This range of cross-national variation suggests
that parental employment patterns and state intervention into childcare provision,
among other factors, heavily influence the demand for grandparental childcare.
Where the principal respondent is the older person themselves, thus from the
grandparent’s perspective, a common question in surveys is to ask:
Do you look after grandchildren?
(Adapted from SHARE, first year of data collection)
70
60
Percentage provision
50
40
30
20
10
0
Greece
Belgium
Austria
Netherlands
Portugal
Germany
Ireland
Great Britain
Spain
Total
Finland
Denmark
France
Sweden
Country
Fig. 9.1 Childcare provided by grandparents by country

(Sources: ESS and calculations of Lewis et al., 2008)
From this question, it is found that many grandparents across Europe look after
their grandchildren; grandfathers and grandmothers alike (Table 9.3). The preva-
lence of grandparental childcare by this measure varies across countries and across
ages, with peaks between 60–65 years. Women are more likely to care for grand-
children than men (Table 6.3). In particular, younger grandparents display notice-
able gender differences in the patterns of caring for grandchildren, with grand-
mothers taking the lion share (Croda and Gonzalez-Chapela, 2005 – not shown in
table).
Hours spent looking after grandchildren are obtained by asking, firstly, how fre-
quent these activities are, and then the number of usual hours spent on the activity
in a given period of time. On average, Europeans over fifty years of age devote
Table 9.3 Prevalence of caring for grandchildren by gender and country
SE DK DE NL FR CH AT IT ES GR
Grandmothers 34.7 39.8 24.7 34.9 35.0 21.5 26.9 25.7 27.1 25.7
(1.5) (1.8) (1.3) (1.5) (1.7) (1.9) (1.3) (1.9) (1.7) (1.4)
Grandfathers 24.4 31.1 20.7 32.0 28.9 17.3 24.7 18.8 20.5 18.7
(1.4) (1.8) (1.2) (1.5) (1.8) (1.8) (1.5) (1.7) (1.7) (1.4)
All grandparents 30.0 35.8 22.9 33.6 32.2 19.5 25.9 22.7 24.3 22.5
(1.1) (1.3) (0.9) (1.1) (1.3) (1.3) (1.3) (1.0) (1.2) (1.0)
Sample of all older persons is 17,629 un-weighted observations using Version 0 release of SHARE
data. All figures are weighted. Figures refer to whole population of older persons having grand-
children. Standard errors are in parentheses.
Source: Adapted from Croda and Gonzalez-Chapela (2005, Table 5A.17).
186 A.S. Koslowski
half an hour a day to looking after grandchildren. A clear geographical gradient is

observed for the amount of time devoted to looking after grandchildren: the average
amount increases as we move south (Croda and Gonzalez-Chapela, 2005). Data
from the Multinational Time Use Study, between 1987 and 1992, have also been
used to calculate average grandparental time. For nine countries, Austria, Canada,
Finland, Germany, Italy, the Netherlands, Sweden, the United Kingdom, and the
United States, time devoted to childcare has been found to be very small, around
0.1 to 0.2 hours per day, tending to be slightly higher for women than for men,
and for younger (in this case aged 45–54) rather than older women (Gauthier and
Smeeding, 2003). This result may be likely to be partly related to the mode of data
collection. Childcare may be performed on an irregular basis and so it may be that
diary surveys, which typically cover only one randomly selected day, are not best
suited to capture a non-daily activity. It may also be the result of an averaging effect
across a heterogeneous population.
To the contrary, the 1998 British Social Attitudes (BSA) study asked all sam-
ple members some broad questions about the role of grandparents and discov-
ered a remarkable level of homogeneity across the sample. “Manifestly, there is
a grand-parenting culture which is shared by the majority of people in British
society” (Dench and Ogg, 2002, p. 13). In the BSA study 35% of grandparents
reported often spending the day and overnight (38%) with their grandchildren with-
out the parents. Only 28% had not spent such a day within the past year (Dench
and Ogg, 2002, p. 22). Grandparents with grandchildren under the age of 12 were
also asked about their role in childcare. Twenty-two per cent reported helping with
daytime childcare on a weekly basis or more. Fifty-three per cent helped with day-
time childcare at least weekly. Forty-two per cent helped with evening childcare
at least monthly. Eighteen per cent took a grandchild to school or nursery at least
monthly.
In summary, the available data on grandparental care from the perspective of the
older person can be disappointingly lacking in detail and fail to highlight the intense
activity identified by studies from the perspective of the grandchild or parent. Whilst
there is a degree of harmonisation across some studies, cross-survey comparison
may be beleaguered by variations in question wording and sample nature. How-
ever, a picture of grandparental care is beginning to be built up. Many grandparents
have regular contact with their grandchildren, but this does not necessarily equate to
substantial amounts of grandparental childcare time. However, a large minority are
heavily involved as the main informal childcare providers, most often in relation to
enabling the labour market participation of the mothers.
Conclusion
A review of current literature confirms that in most West European countries, grand-
parents are a very important source of childcare. Whilst some grandparents are
not involved in grandparental care a substantial minority do provide considerable
amounts of early-years childcare provision. Cross-national variation in the extent to
which grandparents are involved in such informal childcare provision suggests that
there is much that governments can do should their aim be to aid grandparents and
parents.
Despite the current emphasis of EU-level policy on enhancing the formal pro-
vision of childcare to facilitate the labour market activation of mothers as part
of the EU-Lisbon strategy, there is very little evidence of the care activation of
fathers. Partly as a consequence, the role of grandparents in providing informal
childcare thus remains important in all but the Nordic countries (Lewis et al.,
2008). Evidence from the birth cohort studies and the UK Time Use study sug-
gests that reliance on grandparental help may also be a function of the social class
of the parents. This suggests that formal private childcare provision may, in effect,
price certain mothers out of the market, forcing them to find alternative sources of
childcare.
That grandparents are a very important source of childcare is potentially in con-
tradiction with EU policy on ‘active ageing’ (EC, 2002). The amount of hours spent
looking after grandchildren are such that if it were not provided by grandparents
but rather bought in the marketplace, the economic value of this activity would be
considerable (Croda and Gonzalez-Chapela, 2005). Furthermore, if grandparents
increasingly engage with the labour market, then the demand for childcare provi-
sion would most likely further outstrip supply. As such, the role of grandparents as
care providers warrants more recognition in policy development.
It is important to move away from the stereotypical grandparent image of a very
old, inactive person. Many grandparents will themselves be juggling employment
and family commitments. Others may just have settled into what they hope will be
an active retirement involving leisure pursuits, travel and volunteering work. There
is much heterogeneity amongst the population of grandparents, which remains
largely unexplored. Studies of ageing fail to capture data on younger grandparents.
Birth cohort studies fail to capture contextual information about the grandparents.
More reflection is called for as to how best to obtain data on grandparental activity
across the population.
Grandparental time spent on childcare is gendered. Grandmothers are more likely
than grandfathers to be providing informal childcare provision. When grandparents
provide childcare, this has implications for their financial circumstances as well as
for their economic planning for later life. Further work is needed to estimate the
financial costs to the older person of grandparental childcare. It is likely that the
lifetime disadvantage with regard to earnings of informal care is compounded if
women are involved in childcare not once, but twice, and possibly thrice over the
life course.
For both theoretical and policy reasons, it is important to deepen our understand-
ing of this population of grandparental carers and the contribution they make to
informal childcare. Many grandparents may be willing to provide care for their
grandchildren, but this willingness is unlikely to be open-ended. Furthermore, it
may not be sustainable (Dench et al., 1999). More research is called for into the
informal care provision by grandparents for grandchildren and implications of this
activity for all involved.
188 A.S. Koslowski
References
Age Concern (2004) The Economic Contribution of Older People, Age Concern England Policy
Unit, London.
Age Concern (2006) Grandparents, Age Concern Policy Position Papers, Age Concern England
Policy Unit, London.
Anderson, S., Bradshaw, P., Cunningham-Burley, S., Hayes, F., Jamieson, L., MacGregor, A.,
Marryat, L. and Wasoff, F. (2007) Growing up in Scotland: Sweep 1 Overview Report, Scottish
Executive, Edinburgh.
Bamford, C., Gregson, B., Farrow, G., Buck, D., Dowswell, T., McNamee, P., Bond, J. and Wright,
K. (1998) Mental and physical frailty in older people: the costs and benefits of informal care,
Ageing and Society, 18(3): 317–354.
BBC3 (2008) Britain s Youngest Grannies, Part of the Seriously Young Season, Film produced by
BBC3 programmes.
Bolin, K., Lindgren, B. and Lundborg, P. (2007) Your next of kin or your own career? Caring and
working among the 50+ of Europe, Tinbergen Institute Discussion Paper, TI 2007-032/3.
Bradshaw, P., Jamieson, L. and Wasoff, F. (2008) Growing up in Scotland Study: Use of Informal
Support by Families with Young Children, Scottish Government, Edinburgh.
Brennan, D. and Cass, B. (2007) Understanding grandparent care: Policy and practice implica-
tions of grandparents as primary carers of their grandchildren in the Australian context, Paper
presented at the CRFR Extended and Extending Families Conference, University of Edinburgh,
27–29 June.
Broad, B. (2004) Kinship care for children in the UK: messages from research, lessons for policy
and practice, European Journal of Social Work, 7(2): 211–227.
Budig, M. and Folbre, N. (2004) Activity, Proximity or Responsibility? Measuring Parental Child-
care Time, Chapter 3 in Folbre, N. and Bittman, M. (eds.) Family Time: The Social Organization
of Childcare, Routledge, London and New York.
Castles, F. (2004) The Future of the Welfare State: Crisis Myths and Crisis Realities, Oxford Uni-
versity Press, Oxford.
Clarke, L. and Cairns, H. (2001) Grandparents and the care of children: the research and evidence,
in Broad, B. (ed.) Kinship Care: The Placement Choice for Children and Young People, Russell
House Publishing, London, pp. 11–20.
Croda, E. and Gonzalez-Chapela, J. (2005) How do European older adults use their time?, in
Börsch-Supan, A. (ed.) Health, Ageing and Retirement in Europe – First results from the Survey
of Health, Ageing and Retirement in Europe, Mannheim Research Institute for the Economics
of Aging, Mannheim, pp. 265–271.
Crompton, R. (1999), Restructuring Gender Relations and Employment: The Decline of the Male
Breadwinner, Oxford University Press, Oxford.
Crouch, C. (1999) Social Change in Western Europe, Oxford University Press, Oxford.
Daly, M. (2002) A fine balance: Women’s Labor Market Participation in International Comparison,
in Scharpf, F.W. and Schmidt, V.A. (eds.) Welfare and Work in the Open Economy, Vol. II,
Oxford University Press, Oxford, pp. 467–510.
Dench, G and Ogg, J. (2002) Grand-parenting in Britain: A Baseline Study, Institute of Community
Studies, London.
Dench, G., Ogg, J. and Thompson, K. (1999) The role of grandparents’, in Jowell, R., Curtice, J.,
Park, A. Brook, L. and Thomson, K. (eds.) British Social Attitudes: The 16th Report, Ashgate,
Aldershot, pp. 135–157.
Ditch, J., Bradshaw, J. and Eardley, T. (1996) Developments in National Family Policies in 1994,
Social Policy Research Unit. York.
Douglas, G. and Ferguson, N. (2003) The role of grandparents in divorced families, International
EC (1990) 1992 – The Social Dimension. 4th Edition. European Documentation 2/1990, European
Commission, Brussels.
EC (2002) Increasing labour force participation and promoting active ageing, COM (2002) 9 final,
European Commission, Brussels.
Esping-Andersen, G., Gallie, D., Hemerijck, A. and Myles, J. (2002) Why We Need a New Welfare
State, Oxford University Press, Oxford.
Gauthier, A. and Smeeding, T. (2003) Time use at older ages: cross-national differences, Research
on Aging, 25(3): 247–274.
Gershuny, J. (2004) Costs and benefits of time sampling methodologies, Social Indicators
Research, 67: 247–252.
Gornick, J. and Meyers, M. (2003) Families that Work: Policies for Reconciling Parenthood and
Employment, Russell Sage Foundation, New York.
Grandparents plus and Adfam, (2006) Forgotten Families: The Needs and Experiences of
Grandparents who Care for Children whose Parents Misuse Drugs and Alcohol, Adfam and
Grandparents plus, London.
Gray, A. (2005) The changing availability of grandparents as carers and its implications for child-
care policy in the UK, Journal of Social Policy, 34(4): 557–577.
Greengoss, Baroness (4 June 2007) Lords Debate, Pensions Bill, Hansard.
Hank, K. and Buber, I. (2007) Grandparents caring for their grandchildren: findings from the
2004 Survey of Health, Ageing and Retirement in Europe, MEA Working Paper 127–2007,
Mannheim Research Institute for the Economics of Aging, University of Mannheim.
Hansen, K. and Hawkes, D. (2009) Early childcare and child development, Journal of Social Policy,
38(2): forthcoming.
Hirshorn, B.A. (1998) Grandparents as caregivers, in Szinovacz, M.E. (ed.) Handbook on Grand-
parenthood, Greenwood Press, Westport, pp. 200–216.
Hook, J. (2006) Care in context: men’s unpaid work in 20 countries, 1965–2003, American Socio-
logical Review, 71(4): 639–660.
Joshi, H., Paci, P. and Waldfogel, J. (1999) The wages of motherhood: better or worse, Cambridge
Journal of Economics, 23: 543–564.
Juster, F.T. and Stafford, F.P. (1991) The allocation of time: empirical findings, behavioral models
and problems of measurement, Journal of Economic Literature, 29: 471–522.
Kemp, C. (2003) The social and demographic contours of contemporary grandparenthood: map-
ping patterns in Canada and the United States, Journal of Comparative Family Studies, 34(2):
187–212
Land, H. (2002) Spheres of care in the UK: separate and unequal, Critical Social Policy, 22(1):
13–32.
Le Bihan, B. and Martin, C. (2004) Atypical working hours: consequences for childcare arrange-
ments, Social Policy and Administration, 38(6): 565–590.
Lee, Y.-S. (2005) Measuring the gender gap in household labor: accurately estimating wives’ and
husbands’ contributions, in Schneider, B. and Waite, L.J. (eds.) Being Together, Working Apart:
Dual-Career Families and the Work-Life Balance, Cambridge University Press, Cambridge,
pp. 229–247.
Lewis, J., Campbell, M. and Huerta, C. (2008) Patterns of paid and unpaid work in Western Europe:
gender, commodification, preferences and the implications for policy, Journal of European
Social Policy, 18(1): 21–37.
Kroll, B. (2007) A family affair? Kinship care and parental substance misuse: some dilemmas
explored, Child and Family Social Work, 12: 84–93.
OECD (2004) Babies and Bosses – Reconciling Work and Family Life, Volume 3: New Zealand,
Portugal and Switzerland, OECD, Paris.
Price, D. (2006) Why are older women in the UK poor? Quality in Ageing, 7(2): 23–32.
MacKenzie, P. (2007) The multiple layers of mothering work: grandparents raising grandchil-
dren, Paper presented at the CRFR Extended and Extending Families Conference, University
of Edinburgh, 27–29 June.
Murphy, M. and Grundy, E. (2003) Mothers with living children and children with living mothers:
the role of fertility and mortality in the period 1911–2050, Population Trends, 112: 34–44.
190 A.S. Koslowski
Neilson, J. (1998) European briefing: equal opportunities for women in the European Union: suc-
cess or failure, Journal of European Social Policy, 8(1): 64–79.
Richards, A. (2001) Second Time Around: A Survey of Grandparents Raising Their Grandchildren,
Family Rights Group, London.
Rowlatt, J. (2007) Current Affairs, BBC – The One Show, URL www.bbc.co.uk/ theoneshow/
article/2007/07/jr˙grannynannies.shtml accessed 5 May 2008.
Smith, A. and Williams, D. (2007) Father friendly legislation and paternal time across Western
Europe, Journal of Comparative Policy Analysis, 9(3): 175–192.
Smith Koslowski, A. (2008) Who Cares? European Fathers and the Time they Spend Looking after
Children, VDM Verlag Dr Müller, Saarbrücken.
Szikra, D. (2008) Personal correspondence with the author, April.
Wallace, C. (2002) Household strategies: their conceptual relevance and analytical scope in social
research, Sociology: 36 (2) 275–292.
Wheelock, J. and Jones, K. (2002) Grandparents are the next best thing: informal childcare for
working parents in urban Britain, Journal of Social Policy, 31(3): 441–63.
Chapter 10
Internal Migration and Inter-household
Relationships
Oliver Duke-Williams
Introduction
Internal migration is an important agent of population change within the United

Kingdom; it can alter the characteristics of some areas – the balance of different
sorts of people who live there – quite significantly over a short period of time.
Like migration, commuting behaviour also causes massive changes to the popu-
lation of an area on a temporary (usually daily) basis. Both these agents of change,
migration and commuting, are closely linked. Change in usual residential location
inevitably means a change in commuting patterns for those migrants in employment;
people may commute to a new job – possibly by a different mode of transport –
or they may make a commute over a different, possibly longer, route to the same
workplace.
Although we associate migration with flows of people, most research has tended
to consider flows of individuals. A relatively little explored aspect of migration is
the groups of people who move together, sometimes in groups, most frequently as
a household or a family. Moreover, whilst many people move to a new house or
flat in which they are the only occupants, others move into households in which
some people are already living. There is clearly a relationship between group size
(and, perhaps, group behaviour) and household type and the various life stages of
the people in those households. A single person must, by definition, move on their
own, although he or she may move to form a new relationship in a multi-person
household. Young children are likely to move as a part of a family group with their
parent/s, whilst older children may move out of the parental household on their own.
Indeed, mobility is a key driver in household changes, as people move and create
new households.
O. Duke-Williams (B)
School of Geography, University of Leeds, Leeds LS2 9JT, UK
e-mail: o.w.duke–williams@leeds.ac.uk

192 O. Duke-Williams
Much research has focused on the phenomenon of young people leaving home
(Buck and Scott, 1993; Clark and Mulder, 2000; Ermisch, 1999), a process which
is notable for being highly variable in nature in different countries (Aassve et al.,
2002). A sub-group of these migrants are students who leave home as individuals
to live together in large groups in halls of residence or with small groups in pri-
vate households (Duke-Williams, forthcoming). Interest has also focused on shift-
ing attitudes towards and age-specific propensities of marriage and cohabitation
(Berrington and Diamond, 2000), and the role of separation and divorce in residen-
tial mobility (Grundy, 1992). A further avenue for research has been the mobility of
the elderly (Glaser and Grundy, 1998).
In this chapter, groups of migrants who move together are considered in terms
of the general spatial mobility patterns, and the motivations for migration expressed
by persons in different groups and in different types of household. The census data
are used to study overall mobility patterns for groups of different sizes and types.
A feature of migration data from the 2001 Census was the disaggregation by group
type and size, resulting in outputs that were often hard to interpret. One reason for
studying differences between groups is to determine whether a disaggregation of
this sort is always necessary.
The chapter commences by outlining two data sources that are used in the anal-
ysis: the 2001 Census Special Migration Statistics (SMS) and the British House-
hold Panel Study (BHPS). Data from the 2001 Census are then used to examine
movement patterns by migrants in groups of different sizes and to consider whether
there is anything distinctive in the difference between wholly moving households
(where all persons in a household are migrants and had the address one year
before the census) and migrants in other groups. A series of connectivity indices
are produced for migrants in groups of different sizes and of different types, for
both origin and destination specific flows. Correlation coefficients are then exam-
ined in order to investigate the similarity in connectivity levels for all districts for
different groups. Connectivity indices are used as a simple metric for comparing
overall levels of migration in a system. They have the advantage that no popula-
tions at risk (PAR) are required to produce a summary statistic. This is significant,
because the PAR for group level migration are unclear: who is at risk of becoming
a singleton migrant? All people? People who live on their own? There is no clear
answer.
Having examined patterns at the group level, the focus is then changed to
the household level. Data from the BHPS – a panel study – are used to com-
pare expectations of migration with actual behaviour in two ways. Firstly, the
actual process of migration is considered: where people expressed some degree
of desire to migrate, did they actually do so in the following year? Secondly,
for those people that did migrate, what were their reasons for doing so? This
analysis is disaggregated on a household type basis: one might expect people
in different household types to have different reasons for moving, but is this
actually the case? Finally, a comparison is made of spatial and social mobility.
Recording of social class of migrants allows a determination to be made of the
degree to which migrant’s social status changes around the same time that they
move.
10 Internal Migration and Inter-household Relationships 193
Data Sources
Data from the 2001 Census interaction data sets have been used to explore some
general patterns of the way that people move around the country. The 2001 Census
interaction data continued a series of similar data from earlier censuses, but are
collectively larger and more complex than those that preceded them (Rees et al.,
2002; Cole et al., 2002). Three sets were produced – the Special Migration Statistics
(SMS), the Special Workplace Statistics (SWS) and the Special Travel Statistics
(STS) – at three spatial scales. Whereas the SMS provide UK-wide spatial coverage,
the other two sets are spatially constrained. The STS were generated for residences
in Scotland only, while the SWS were produced for residences elsewhere in the
UK. The STS are effectively a superset of the SWS for flows with destinations in
Scotland, in that they include both journeys from home to work (as do the SWS)
and journeys from home to a place of study for full-time students (including school
children).
The interaction or origin-destination data were produced at three spatial scales
referred to as levels. Level 1 refers to the local authority districts across Great
Britain – an amalgam of different types of local government authorities in England,
Wales and Scotland – together with parliamentary constituencies in Northern
Ireland. Level 2 involves ward level data and contains an amalgamation of census
area statistics (CAS) wards in England, Wales and Northern Ireland, and standard
table (ST) wards in Scotland. The combinations of component geographies used in
these two levels are distinct to the interaction data, and are respectively referred to
as ‘interaction data districts’ and ‘interaction data wards’. Level 3 spatial units are
output areas, a new geography created for the outputs of the 2001 Census (Martin,
2002). The set of areas used for the interaction data are equivalent to those areas
used for other census outputs.
Across all three sets of interaction data, results are presented in a variable num-
ber of cross-aggregated tables. The number of tables available, and the numbers of
categories used in component variables varies with spatial scale: at level 1 there are
more tables, with finer classifications used and at level 2, there are fewer tables with
coarser classifications (Stillwell and Duke-Williams, 2007). For both SMS and the
SWS/STS, there is only a single table available at level 3, the finest spatial scale: an
age by sex table in the case of the SMS and a mode of travel table in the case of the
SWS and STS.
In contrast to the 2001 Census, the BHPS can be used to explore connections
between household structure and migration events. The BHPS is a panel study that
commenced in 1991 and has been updated with annual survey waves since then.
As a panel study, it follows the same groups of individuals over time. A distinct
feature of the BHPS is that it is household based, with all adult members of survey
households being interviewed, and children aged 11+ also completing a short inter-
view. The survey is a multi-purpose study with core and additional elements cov-
ering many areas including social attitudes, employment and income, housing and
tenure, health, relationship histories, and individual and household demographics.
The survey also includes questions on residential mobility, which are explored in
this chapter.
The initial panel consisted of over 5,000 households containing over 10,000
individuals, and this panel has grown over time through the inclusion of additional
samples in Scotland, Wales and Northern Ireland. The panel accumulates new mem-
bers as a result of mobility: when a sample member moves to a new household, the
members of that household also become sample members. However, mobility is also
a major cause of sample attrition, as panel members cannot always be traced; other
losses include refusal to be interviewed. By wave 13, the total number of persons
interviewed had fallen to 8,655 (Lynn, 2006).
General Patterns of Migration

The 2001 Census interaction data provide useful information about migration and
commuting within the UK. They have particular strengths over sample studies in
that they include the whole population. This is of particular significance with a phe-
nomenon such as migration, because ‘migrants’ are a relatively small sub-set of
the whole population: about 10% of the UK population were internal migrants by
the Census definition in the period 2000–2001. Persons are defined as migrants by
the 2001 Census if they have changed their usual residence in the year prior to the
Census. The propensity to migrate varies with various socio-demographic charac-
teristics (Champion, 2005), the most notable of which is age. Figure 10.1 shows
the numbers of internal migrants in the UK per 1,000 persons, by quinary age
groups. The distribution is a typical one for migration, with the highest rates being
found amongst young adults. Rates are also relatively high for the elderly, and for
350
Male Female
Migrants per 1,000 persons
300
250
200
150
100
50
0
0–4
5–9
10–14
15–19
20–24
25–29
30–34
35–39
40–44
45–49
50–53
55–59
60–64
65–69
70–74
75–79
80–84
85–89
90+
Age group
Fig. 10.1 Internal migration rates by age, UK, 2000–2001

(Source: 2001 Census SMS l; author’s calculations.)
young children. Another characteristic of migration illustrated in Fig. 10.1 is that

unlike many other social phenomena, overall migration propensities in countries
like the UK exhibit little gender difference. Further analysis of age and gender vari-
ations in the 12 months before the 2001 Census is available in Dennett and Stillwell
(forthcoming).
The term ‘migration’ covers a variety of types of mobility, from persons mov-
ing on their own to groups of several persons moving together. In censuses prior to
the 2001 Census, a distinction was drawn between ‘wholly moving households’ and
‘other groups of migrants’. The term wholly moving household refers to a house-
hold in which all persons were migrants and all had the same usual residence one
year prior to the census; the definition encompasses persons who have moved as
singletons. The term does not necessarily mean that all persons who were in the
original household moved – it may be the case that one or more persons remained
in that household, and this might particularly be the case with singletons when, for
example, a young adult has moved out of their parental home and into their own
accommodation. It might be possible to identify truly ‘wholly moving’ cases (that
is, where all persons in the original household have moved together) using original
census data, but such analysis would be complicated and expensive, and probably
prone to error.
Table 10.1 illustrates the changes observed in the number of wholly mov-
ing households in Great Britain in the 1991 and 2001 Censuses. The number of
wholly moving households, and the number of migrants contained within them
both increased when the two censuses are compared. This apparent increase should
be understood in the context of changes in both the census base population, and
the handling of migrants from an unstated origin. The 2001 Census included stu-
dents at their term-time residence rather than at their parental residence, as was
the case in 1991; this may have affected either their household residency status, or
their membership of wholly moving households. In addition, as part of the 2001
Census processing, migrants’ origins that were missing or not fully stated were
imputed, whereas in the 1991 SMS, they were tabulated separately. No migrants
Table 10.1 Numbers of wholly moving households and migrants in Great Britain, 1990–1991 and
2000–2001
1990–1991 2000–2001
Number of migrants in households 4,463,974 5,670,422

Number of wholly moving households 1,366,115 1,665,469
Migrants in wholly moving households 3,085,854 3,451,938
Mean number of migrants per wholly 2.3 2.1
moving household
% of migrants in households 69 61
Sources: 1991 Census SMS Set 2, Table 2; 1991 Census SMSGAPS Table 6; 2001 Census SMS
Level 1, Table 6.
with unstated origins were coded as being in ‘wholly moving households’ in 1991,
whereas migrants in 2001 with imputed origins could have been coded as being in
wholly moving households.
Table 10.1 also shows the mean number of migrants in each wholly mov-
ing household, and the percentage of all migrants in households who were clas-
sified as being in wholly moving households. Despite the apparent increase in
the number of migrants in wholly moving households between 1991 and 2001,
the mean number of migrants per group had fallen, as had the percentage of all
migrants that were in wholly moving households. It is possible that the decrease
in mean group size is a result of significant numbers of migrants with imputed
origins being coded as single person wholly moving households. However, the
distribution of group sizes was not published for 1991 data, so this explanation
remains speculative. The decrease in the proportion of migrants who were in wholly
moving households may be related to improved capture of data relating to stu-
dents, and the growth in the period of the numbers of students living in shared
households.
The 2001 Census processing introduced an additional way of classifying groups
of migrants: that of the moving group. A moving group is a group of persons who
were migrants and had the same usual residence one year prior to the census. A
wholly moving household, using the previous definition, is thus one example of a
type of moving group. However, the classification also includes non-wholly moving
groups; these include the cases where:
• one or more persons in the household is a migrant but not all household members
are; and
• all persons in the household are migrants but do not all have the same usual
residence one year prior to the census.
Identification of moving groups at the time of census processing was exhaus-

tive: that is, all migrants in households were allocated to a moving group.
Table 10.2 shows the national distribution of migrants in households (as opposed
to communal establishments) according to their group status. The ‘one per-
son’ counts are the same for groups and for migrants for both wholly mov-
ing households and other moving groups. There were approximately 5.8 million
persons within the UK in 2000–2001 migrating in 3.5 million moving groups,
of which 48% were wholly moving households and 52% were ‘other mov-
ing groups’. A very high proportion of those persons moving in other moving
groups were in fact individual movers (85%) compared to 42% of wholly moving
households.
Amongst the 3.5 million persons migrating in wholly moving households, over
half involved three or more persons moving together, with 28.3% in two person
households and one fifth as single persons. In contrast, over two thirds of migrants
in other moving groups were single persons, with similar numbers split between two
person and three or more person groups.
Table 10.2 Groups and migrants within the UK, 2000–2001
Number of persons
1 person 2 persons 3+ persons All
Wholly moving households

Groups 719,379 500,461 486,356 1,706,196
Migrants 719,379 1,000,922 1,821,914 3,542,215
Other moving groups
Groups 1,545,286 178,041 104,753 1,828080
Migrants 1,545,286 356,082 367,029 2,268,397
Source: 2001 Census SMS level 1, Table 6.
Migration Connectivity
The results shown in Tables 10.1 and 10.2 indicated that whilst the majority of
migrants are in wholly moving households, there are a very significant number who
are in sub-household size groups. Table 10.1 suggested that the proportion in wholly
moving households may have fallen between 1991 and 2001, although changes to
the population base and the way in which migrants with an unstated origin were pro-
cessed mean that this interpretation must be cautious. An obvious question arises: do
migrants in different sorts of groups (both in terms of size, and in terms of whether
the group formed a wholly moving household or not) have different mobility pat-
terns? If they do, then the distinction offered in the 2001 Census between different
sorts of groups is a useful and important one; if they do not then disaggregation by
group type and size is an unnecessary complication of the output tables that would
serve only to confuse users.
In order to compare the characteristics of migrants at a national scale in dif-
ferent groups, connectivity indices (Bell et al., 2002) were constructed for wholly
moving groups, and for other groups which did not form a wholly moving house-
hold. Index values were calculated for all groups at the district level. A connectivity
index is a simple measure that shows the extent to which any one place is con-
nected to other places through migration. If the index for any location is equal
to 1.0, then that place is connected to all other places in the universe of inter-
est (that is, it has a non-zero observed flow to or from all other locations); if the
index has a value of 0.0 then that location is connected to no other places. Sep-
arate sets of indices were constructed from an origin perspective and from a des-
tination perspective. A high origin-specific index means that the location sends
migrants to a large number of other locations, and a high destination-specific
index indicates that the location receives migrants from a large number of other
locations. The index value is dependent on the spatial scale for which it is con-
structed: the coarser the geography, the higher the index values will be. However,
given a fixed geographic scale, it provides a simple mechanism for comparing flow
characteristics.
One person Two person Three or more

household household person household
0.00 – 0.13 0.00 – 0.13 0.00 – 0.13
0.14 – 0.19 0.14 – 0.19 0.14 – 0.19
0.20 – 0.27 0.20 – 0.27 0.20 – 0.27
0.28 – 0.00 0.28 – 0.00 0.28 – 0.00
Fig. 10.2 Migration destination connectivity index values for wholly moving households;
2000–2001 (a) Single person households (b) Two person households (c) Three or more person
households
(Source: Author’s analysis based on census data.)
Figure 10.2 shows the values of the destination-specific connectivity index for
wholly moving households consisting of (a) one person, (b) two people and (c)
three or more persons. The shading scheme is different for each map, and in each
case is based on result quartiles: the darkest shaded areas are the 25% of dis-
tricts with the highest destination-specific connectivity values. In absolute terms,
there is a steady decline in connectivity as the group size increases: connectivi-
ties are higher for single person households than for two person households, and
in turn, connectivities are higher for two person households than for three person
households. Two features are apparent in this set of maps. Firstly, districts in Scot-
land have high connectivity values, and secondly, the spatial pattern elsewhere is
variable.
The high levels for Scottish Council Areas are in part an artefact of the sta-
tistical disclosure control methods applied to the interaction (and other census)
data, although some Scottish destinations are popular of course and thus well con-
nected to other locations. The data were subject to a statistical disclosure control
method termed SCAM, the small cell adjustment methodology (Duke-Williams and
Stillwell, 2007). Small values (the definition of small was not disclosed by the cen-
sus agencies but was widely understood to be the values 1 and 2) in the original
data were modified to become either value 0 or 3. However, this procedure was
not followed in Scotland, including those parts of the SMS that related to destina-
tions in Scotland. Thus, a significant number of small flows (which might otherwise
have been changed to 0) remained present in the data relating to Scotland, with
the result that more distinct flows were observed in the published data for Scottish
destinations.
The patterns of destination connectivity outside Scotland are not affected in this
way, and are thus of more direct interest. The maps shown in Fig. 10.2 show a shift-
ing pattern of destination preference given group size. A good example of this is
for London and the surrounding area – this has high connectivity values for sin-
gle person households: people who were living on their own at the time of the
2001 Census, and had been migrants in the preceding year, had moved there from
a wide range of origin districts. This is less marked for two person households, and
there is a low connectivity ‘hole’ around London for wholly moving households
of three or more persons. This finding indicates that people in larger wholly mov-
ing households moved to the London area from a limited set of origins; it does
not necessarily indicate a difference in the absolute numbers of households or per-
sons involved. Another visual effect that is clear on the map is that the darkest
shaded areas appear to take up more space for the larger groups. This is because the
areas in the highest connectivity quartile are larger less urban districts. The inverse
of this is less immediately visually obvious: that for single person households,
those districts with the highest connectivity index values tend to be small urban
districts.
In order to determine the extent to which flow patterns differed by group size, by
wholly or partly moving status, and by origin or destination specificity, correlation
coefficients were calculated for various sets of migration connectivity indices. A
high coefficient indicates that the pairs of connectivity indices are similar: that each
zone under consideration dispatches (or receives) migrants from a similar number
of areas. Thus, if the connectivity indices have a high correlation, those areas in
the first group that are highly connected are also highly connected in the second
group, whilst those areas that are poorly connected in the first group are also poorly
connected in the second group. It does not necessarily mean that each pair of areas
being compared are connected to the same partner areas.
Table 10.3 shows the correlation coefficients for each pair of origin-specific
migration connectivity indices. Comparisons are made for all groups (i.e., one per-
son against two persons, one person against three plus persons and two persons
versus three plus persons) within the wholly and partly moving sets, and for each
same-size group where wholly and partly moving groups are compared. Compar-
isons within each ‘family’ (‘wholly moving households’ or ‘other groups’) explore
the effect of group size on behaviour, accounting for any wholly or partly moving
differences; comparisons between these families explore the effect of wholly ver-
sus partly moving status whilst accounting for group size. The correlations for the
origin-specific values are lowest (suggesting different patterns of movement) within
the sets of ‘other groups’ – those who did not form wholly moving households. This
suggests a difference in origin connectivity for these ‘other groups’: areas that are
highly connected (that is, distribute migrants to a large number of destinations) for
single persons, are not necessarily highly connected for larger groups. This differ-
ence is most marked when single person groups are compared with groups of three
Table 10.3 Comparison of correlation coefficients for origin-specific values
Wholly moving
households Other groups
1 2 3+ 1 2 3+
Wholly moving 1 person – 0.94 0.88 0.94 – –

households 2 persons – 0.93 – 0.90 –
3+ persons – – – 0.87
Other groups 1 person – 0.87 0.74
2 persons – 0.88
3+ persons –
Source: Author’s analysis based on census data.
or more persons. The indices are more strongly correlated for wholly moving house-
holds: areas that are strongly connected for single persons also tend to be strongly
connected for larger groups.
Equivalent results are shown for destination-specific connectivity indices in
Table 10.4. Almost all coefficient values are lower for the destination-specific index
comparisons than for the equivalent comparisons amongst origin-specific values.
This suggests greater variation by group size and by wholly/partly moving status
when migration is viewed from the destination perspective – areas that are strongly
connected as destinations (that is, they receive migrants from many other districts)
are not necessarily strongly connected if the group size or type changes. This is
most marked for the comparison of wholly moving households, in particular for
single person households in comparison to households of three or more persons.
The final set of correlation coefficients are shown in Table 10.5. These show
the comparisons of origin versus destination-specific values for fixed group size
and group type. The coefficient values decrease with group size for both group
types, indicating that the larger the group size, the greater the difference in origin
and destination-specific patterns. Given that the origin-specific values have a strong
within-type correlation, this suggests that destination choice is more selective as
group size increases.
Table 10.4 Comparison of correlation coefficients for destination specific values
Wholly moving
households Other groups
1 2 3+ 1 2 3+
Wholly moving 1 person – 0.74 0.64 0.90 – –

households 2 persons – 0.86 – 0.67 –
3+ persons – – – 0.73
Other groups 1 person – 0.88 0.70
2 persons – 0.81
3+ persons –

Table 10.5 Comparison of correlation coefficients for origin and destination specific values
Destination-specific
Wholly moving households Other groups
Origin-specific 1 2 3+ 1 2 3+
Wholly moving households 1 person 0.86 – – – – –

2 persons 0.58 – – – –
3+ persons 0.52 – – –
Other groups 1 person 0.93 – –
2 persons 0.87 –
3+ persons 0.69
The importance of these correlation coefficients lies in the degree to which they
indicate that connectivity patterns are not correlated when two migrant populations
are compared. If two populations (either different group sizes, or different group
types) have a high correlation, it suggests that their mobility patterns are similar,
and the more similar they are, the less need there might be for separating results by
group type. However, where the correlation is low, the results suggest that there
are differences (in connectivity at least) by group type or size. Tables 10.3 and
10.4 explored this issue. Coefficients were lower for destination-specific indices
than for origin-specific indices, indicating that group size was important when con-
sidered from this perspective. Table 10.5 summarised differences when origin and
destination-specific indices were compared. Correlation coefficients were lower for
wholly moving households and also decreased as group size increased. These results
indicate that out-migration is not a reliable indicator of in-migration, and further that
this becomes more true for larger groups of migrants.
Reasons for Moving
The previous section of this chapter used the 2001 SMS to examine differences in
migration behaviour by group size, group type and origin/destination perspective.
Whilst the census data are spatially detailed and refer to a large sub-population,
they are limited with respect to migration analysis in that they do not offer any
direct information about people’s status prior to migration, or for the reasons that
they might have had for migrating. In order to investigate these aspects, the BHPS
data were used.
The BHPS consists of numerous waves of data. Due to sample attrition, the
sample size has tended to fall with each wave. Waves J and K (the tenth and
eleventh waves) were selected for analysis despite this attritional aspect, because
they roughly match the transition period over which the census identified migration.
Wave J was collected in the period 2000–2001, and wave K collected a year later.
An initial comparison of data in wave J with results from the 2001 Census allows
the general characteristics of the data sets to be compared. Figure 10.3 shows the age
80
70
Migration frequency 60
50
40
30
20
10
0
0 20 40 60 80 100
Age
Fig. 10.3 Age profile of BHPS migrants

(Source: BHPS wave J; author’s calculations.)
profile of migrants recorded in the BHPS, identified as movers within the past 12
months at the time of interview. Owing to the nature of the BHPS sample, children
are not included in this graph and, for other age groups, the data show a similar
distribution to those of migrants observed in the 2001 Census (Fig. 10.1) but for
individual ages rather than quinary age groups. The data for wave J show interesting
variations within quinary age groups but, although they include the region of former
residence for migrants as well as the region of current residence, they are too limited
to generate a usable migration matrix which might be compared to the equivalent
matrix from the census data.
One interesting aspect of the BHPS data is that all persons – not just recent
migrants – are asked about the date at which they moved to their current residence.
The results of this question are shown in Fig. 10.4 for (a) all migrants in the data
set and (b) those migrants who moved to their present residence in the last 30 years.
Figure 10.4a is dominated by a large spike of recent migrants, and thus in Fig. 10.4b
the number of migrants is shown using a logarithmic scale in order to make trends
more visible. The graphs are dominated by recent moves. This is to be expected,
as the question refers only to the most recent change of residence; those who have
changed residence have probably done so recently, although this observation does
not indicate when any other moves had previously been made.
The dominance of recent migration events means that comparisons of migration
behaviour between the BHPS and the 2001 Census have a reasonable degree of
validity: both data sets are focused on recent migrants.
Expectation of Moving
The BHPS survey includes two prospective questions about mobility: respondents
are asked whether they would prefer to move away from their current residence,
1600
All migrants
Number of sample persons

1400
1200
1000
800
600
400
200
0
1930 1940 1950 1960 1970 1980 1990 2000
Year moved to present address
10,000
Detail of 1970 onwards, log scale
Number of sample persons
1000
100
10
1
1970 1975 1980 1985 1990 1995 2000
Year moved to present address
Fig. 10.4 Year moved to present address by BHPS migrants (a) All migrants (b) detail 1970
onwards, log scale
(Source: BHPS wave J; author’s calculations.)
and also whether they expect to move within the next 12 months. Table 10.6 shows
the separate and joint responses for these two questions. Of all persons with a valid
answer, 33% indicated that they would prefer to move house. However, there are
obvious constraints to mobility including occupational, domestic and social ties as
well as financial considerations which mean that a general desire does not necessar-
ily mean that a person really intends to move. Of the same respondents, 14% said
that they expected to move within the next 12 months. The main part of Table 10.6
shows the expectations of persons with different preferences: of all those who pre-
ferred to move, 64% did not in fact expect to move. There was closer correlation, as
might be expected, for non-movers: of those who preferred to stay where they were,
91% did not expect to move within the next 12 months.
Preference for a move varies by many characteristics, including age and house-
hold type. Table 10.7 summarises the proportions of persons expressing a prefer-
ence for moving by gender and household type, for four age groups. In each row,
Table 10.6 Preference for and expectation of moving
Expect to move
Prefer to move? Don’t know (4%) Yes (14%) No (82%) Total (100%)
Don’t know (1%) 19% 18% 63% 191

Stay here (66%) 2% 6% 91% 9,915
Prefer to move (33%) 8% 29% 64% 4,975
Total 641 2,073 13,367 15,081
Source: BHPS wave J; author’s calculations.
Table 10.7 Preference for moving
Age group
Prefer to stay All ages 16–24 25–44 45–64 65+
% % % %
By gender
Male 7,177 57 54 66 78
Females 8,426 54 58 69 79
By household type
Single non-elderly∗ 1,085 55 50 62 67
Single elderly∗ 1,237 – – 75 84
Couple: no children 4,406 53 – 73 79
Couple with dependent children 4,563 60 60 70 90
Couple with non-dependent children 1,839 57 56 71 82
Lone parent with dependent children 795 53 52 57 64
Lone parent with non-dependent children 536 66 50 69 85
2+ unrelated adults 356 58 54 52 80
Other households 262 47 50 68 72
∗ - ‘Elderly’ is gender-specific retirement age. A small number of persons are mis-coded as
elderly when aged 64, and as non-elderly when aged 65.
Source: BHPS wave J; author’s calculations.
the table shows the percentage of persons in that category (male, female, et cetera)
of that age group, who indicated a preference to stay. All persons in all households
are included in this comparison. An ‘all ages’ base is shown for each category, indi-
cating that the numbers of persons in each age group vary by category. In all cases,
the strongest preference for staying is found amongst the oldest age group, and in
most cases the lowest preference for staying is found in the age group 25–44. The
highest preference for staying across all ages is found for persons in a couple family
with dependent children.
Realisation of Expectation
The results described above were obtained by studying a single wave of the BHPS
data. By linking more than one wave of panel data, it is possible to study changes
Table 10.8 BHPS: Individual mover status at wave K by preference expressed at wave J
Individual mover within GB status

Preference Non-mover (%) Mover (%) Total
Don’t know 86 14 164

Stay here 95 5 9,068
Prefer to move 82 18 4,395
Total 91 9 13,627
Source: BHPS waves K and J; author’s calculations.
that occur in the lives of respondents. The BHPS includes both prospective questions
about migration – described above – and retrospective questions: did people move,
and if so why did they do so, et cetera. This combination permits the investigation of
the extent to which intentions of migration are realised in practice. In order to exam-
ine this issue, wave K was joined to wave J through the available linking variables.
The resulting records were a subset: those persons present in wave J who were also
present in wave K.
Having linked the data, the observed responses for migration preference and
expectation at wave J were cross-tabulated with the individual move status of the
same individuals when interviewed at wave K. Table 10.8 shows the actual migra-
tion behaviour of individuals in the 12 months before being interviewed for wave
K, given the preferences that they expressed a year earlier in wave J. The results are
broadly similar to the cross-tabulation of preference versus expectation for wave J
in isolation. Of those who preferred to move, 82% were in fact non-movers. When
wave J was considered in isolation, 64% of those who preferred to move did not
expect to move. Of those persons who preferred to stay where they were, 95% were
non-movers in practice: this closely matches the figure of 91% in wave J for those
who preferred to stay where they were, and did not expect to move.
It might be assumed that those who said that they expected to move were more
likely to do so. Table 10.9 shows the actual migration behaviour of individuals in
wave K, given the expectation of doing so that they expressed a year earlier in wave
J. Those who did not expect to move overwhelmingly did not: 97% were non-
movers. Of those who said that they did expect to move, only 47% actually were
movers. This is a much larger translation from expectation to realisation than was
the case for those that simply had a preference for moving, but indicates that simple
expectation of movement is not a hugely reliable indicator of actual behaviour.
It is possible that the comparison of mover status for those that expected to move
will under-estimate the true conversion of expectation to realisation of migration,
due to unforeseen delays in the process of moving – problems in selling or buying a
property, or the need to wait for a tenancy agreement to expire. In order to examine
whether there was a deferred conversion from expectation to realisation, a third
wave – wave L – was further linked to the combined data set, and the mover status
for persons at wave L were considered in the context of expectations expressed at
wave J. Table 10.10 shows the results of this comparison. Of those who stated in
Table 10.9 BHPS: Individual mover status at wave K by expectation expressed at wave J
Individual mover status

Expectation Non-mover (%) Mover within GB (%) Total

Yes 53 47 1,679
No 97 3 11,387
Total 91 9 13,627
Table 10.10 BHPS: Individual mover status at wave L by expectation expressed at wave J
Individual mover status

Expectation Non-mover (%) Mover within GB (%) Total

Yes 73 27 696
No 95 5 9,101
Total 93 7 10,182
Source: BHPS waves K and L; author’s calculations.
wave J that they expected to move within the next 12 months, 27% moved in the
period between waves K and L, roughly 12–24 months after the initial interview.
Some of these people may also have moved in the period between waves J and K:
a three way comparison was not done due to the overall limited numbers of persons
involved. Assuming that a majority of the movers between K and L had not also
been movers between J and K, it appears that there was a limited degree of deferred
realisation of the initial expectation, although probably enough at least to state that
a majority of those who expected to move had completed a move within the two
years following wave J.
Reasons for Moving
As well as including questions about intentions to move and actual moves, the BHPS
survey also includes both prospective and retrospective questions about reasons for
moving. Those individuals who stated that they would like to move are routed to
additional questions about the reasons for this, and those individuals who indicate
that they have moved are similarly routed to questions asking about the reasons for
the move. Through linking successive waves of data, it was possible to compare
stated reasons for moving (for those that did move) for two different groups: those
who expected to move and did so, and those who did not expect to move, but in
practice did so. The second group will include people who were in some way forced
into moving for some unexpected reasons. An initial question in the survey routing
asked whether people moved for employment reasons; for those who answered ‘no’
Table 10.11 Stated reasons for moving, by household type
Household type Most commonly stated reason % within h’hold type H’hold type base
Single: non-elderly Move to college 17 160

Single: elderly Health reasons 22 37
Couple: no children Move in with partner 20 340
Couple with Larger accommodation 28 364
dependent children
Couple with Smaller accommodation 15 41
non-dependent
children
Lone parent with Split from partner 25 55
dependent children
Lone parent with Split from partner 20 20
non-dependent
children
2+ Unrelated adults Evicted, repossession 23 73
Other households Other 22 23
Source: BHPS wave K; author’s calculations.
supplementary questions were asked about other reasons. Moves for employment
reasons were low for both groups: 20% of those who had expected to move identified
the reason as employment related, whilst 10% of those who had not expected to
move cited employment reasons. A wide variety of reasons were given including
relationship formation and relationship breakdown, and moves to either larger or
smaller accommodation. There is no obvious pattern to the set of reasons stated
if they are considered simply as set ranked on the basis of how often they were
identified.
One reason for this is that reasons for moving depend strongly on life stage,
and changing occupational and social issues related to life stage. Table 10.11 shows
the most commonly cited reason for moving given by individuals in wave K, dis-
aggregated by household type. The results are perhaps surprisingly stereotypical,
although the degree to which any particular reason dominates varies by household
type. The table only shows the most commonly cited reason: in some cases other
reasons were cited almost as frequently; for example for the household type ‘lone
parent with non-dependent children’, eviction and moving in with family were as
common as ‘split from partner’
It is striking that the most commonly stated reasons fit an assumed model of life-
stage: for a couple with no children, the most commonly stated reason for migration
in the past year was to move in with a partner – presumably a new relationship
formation with the current partner. Couples with dependent children stated that they
have moved in order to gain larger accommodation – presumably as a result of a
growing family, whilst couples with older (non-dependent) children stated that they
moved to smaller accommodation. For lone parents – whether with dependent or
non-dependent children, the most commonly stated reason was due to a split from
a previous partner. Somewhat neatly, for ‘Other households’, the most commonly
stated reason was ‘other’.
Change in Circumstance
A drawback with using census data for migration analysis is that it is a cross-
sectional snapshot, whilst migration is a process enacted over time. Furthermore,
migration is often associated with significant changes in circumstance: a change
in occupation or family structure for example. As a panel study, the BHPS is very
useful for examining changes that have occurred over the same transitional time
frame as used to identify migration events. Table 10.12 shows net changes in pro-
portions of persons in different Registrar General’s social class groups between the
two waves analysed. The table is split into two groups: those who had moved in the
12 months preceding wave K, and those who had not.
The proportionate changes for the movers are larger than for the non-movers,
although in some cases this may be an artefact of the smaller numbers involved
in the migrant group. The overall figures suggest that migration is linked to upward
social mobility: the highest two groups grew in size, whilst others had a net decrease.
Closer examination of this apparent effect can be seen by constructing a social class
transition matrix.
Table 10.13 shows the net changes between social classes for individuals in
waves J and K of the BHPS. Thus, there was a net shift of three persons from
managerial and technical occupations to professional occupations. The largest net
change is from skill non-manual to managerial and technical.
Table 10.12 Change in Registrar General social class between BHPS waves J and K, by mover
status
Movers Non-movers
Social class Wave J Wave K Change (%) Wave J Wave K Change (%)
Professional 46 50 9 351 345 –2

occupations
Managerial & technical 233 252 8 2,137 2,200 3
occupations
Skilled non-manual 177 170 –4 1,636 1,630 0
Skilled manual 146 137 –6 1,358 1,329 –2
Partly skilled 110 105 –5 958 947 –1
occupations
Unskilled occupations 26 25 –4 295 285 –3
Armed forces 1 0 –100 7 6 –14
Source: BHPS waves J and K; author’s calculations.
Conclusion
The research described in this chapter has explored different aspects of group related
migration. If an indicator such as the index of migration connectivity is mapped for
different group types and sizes, there appear to be differences in the patterns shown,
although the extent of these differences does vary. These differences were made
more clear in the set of comparisons of correlation coefficients between different
Table 10.13 Net transitions between social classes by movers between BHPS waves J and K
Managerial & Skilled Partly Unskilled

Professional technical non- Skilled skilled occupa Armed
Movers occupations occupations manual manual occupations tions forces
Professional 0
occupations
Managerial & 3 0
technical
occupations
Skilled 0 17 0
non-manual
Skilled manual 0 6 6 0
Partly skilled 1 –2 2 1 0
occupations
Unskilled 0 1 2 1 –3 0
occupations
Armed forces 0 0 0 0 0 0 0
observations. For origin-specific connectivities, migrants moving as wholly mov-

ing households were closely correlated regardless of group size, whereas migrants
moving in other groups were less closely correlated – suggesting that these ‘other’
groups are more diverse. The same was not true when destination-specific connec-
tivities were compared: results were slightly less closely correlated for wholly mov-
ing groups than for other groups. However, the most significant observation is that
all destination-specific sets of comparisons exhibited lower correlation coefficients
than was the case for the origin-specific comparisons. The relationships between
origin and destination-specific correlations was most marked, with a clear decline
in correlation as group size increases. Thus, for single-person groups, areas sent
migrants to and received them from a similar number of areas, but as group size
increased, this became less common.
The census data were useful in establishing the existence of patterns based on
group size and type, but due to the nature of the census they do not indicate anything
about the reasons for migration. The BHPS data were useful in this regard. There
were clear links in the results between household type, acting as a proxy for life
stage, and the most commonly cited reason for migrating for persons in each house-
hold type. Indeed, the relationship was so stereotypical that Table 10.10 could easily
by re-arranged in chronological order, with parallel paths through couple or sin-
gle parent status. Through being a panel study, individuals’ circumstances could be
compared before and after a migration had occurred. The analysis showed changes
in social class that were associated with migration. It is not easy to establish a causal
link for such changes – Do they come about following, and because of, a migration?
Do they happen first and cause a migration? – and any causal relationship is likely
to vary between groups. A further aspect of census data (or any similar retrospective
transitional data) is that by its nature it only gathers data on successfully completed
migration events. The BHPS data allowed investigation into cases in which people
expressed a wish to migrate, but did not in fact do so. The research demonstrated
and quantified differences in the expectation and subsequent realisation of migra-
tion. Further work in this area – assessing those who did not migrate – may be
useful in investigation of the constraints on migration.
References
Aassve, A., Billari, F., Mazzuco, S. and Ongaro, F. (2002) Leaving home ain’t easy, a comparative
longitudinal analysis of ECHP data, Journal of European Social Policy, 12(4): 259–275
Bell, M., Blake, M., Boyle, P., Duke-Williams, O., Rees, P., Stillwell, J. and Hugo, G. (2002)
Cross–national comparison of internal migration: issues and measures, Journal of the Royal
Statistical Society A, 165(3): 435–464.
Berrington, A. and Diamond, I. (2000) Marriage or cohabitation: a competing risks analysis of first-
partnership formation among the 1958 British Birth Cohort, Journal of the Royal Statistical
Society. Series A, 163(2): 127–151
Buck, N. and Scott, J. (1993) She’s leaving home: but why? An analysis of young people leaving
the parental home, Journal of Marriage and the Family, 55: 863–874.
Champion, A. (2005) Population movement within the UK, Chapter 6 in Chappell, R. (ed.) Focus
on People and Migration, Palgrave Macmillan, Basingstoke, pp. 91–113.
Clark, W. and Mulder, C. (2000) Leaving home and entering the housing market, Environment and
Planning A, 32: 1657–1671.
Cole, K., Frost, M. and Thomas, F. (2002) Workplace data from the census, Chapter 19 in
Rees, P., Martin, D. and Williamson, P. (eds.) The Census Data System, Wiley, Chichester,
pp. 269–280.
Dennett, A. and Stillwell, J. (forthcoming) Age and sex migration patterns at the start of the new
millennium, in Stillwell, J., Duke-Williams, O. and Dennett, A. (eds.) Technologies for Migra-
tion and Commuting Analysis: Spatial Interaction Data Applications, IGI Global, Hershey.
Duke-Williams, O. (forthcoming) The geographies of student migration in the UK, Environment
and Planning A.
Duke-Williams, O. and Stillwell, J. (2007) Investigating the potential effects of small cell
adjustment on interaction data from the 2001 Census, Environment and Planning A, 39(5):
1079–1100.
Ermisch, J. (1999) Prices, parents and young people’s household formation, Journal of Urban
Economics, 45(1): 47–71.
Glaser, K. and Grundy, E. (1998) Migration and household change in the population aged 65 and
over, 1971–1991, International Journal of Population Geography, 4: 323–339.
Grundy, E. (1992) The household dimension in migration research, in Champion, T. and Fielding,
T. (eds.) Migration Processes and Patterns: Vol. 1. Research Progress and Prospects, Belhaven
Press, London.
Lynn, P. (ed.) (2006) Quality Profile: British Household Panel Survey, Version 2.0: Waves 1 to 13:
1991–2003, Institute for Economic and Social Research, London.
Martin, D. (2002) Geography for the 2001 Census in England and Wales, Population Trends, 108:
7–15.
Rees, P., Thomas, F. and Duke-Williams, O. (2002) Migration data from the census, Chapter 18 in
Rees, P., Martin, D. and Williamson, P. (eds.) The Census Data System, Wiley, Chichester, pp.
245–267.
Stillwell, J. and Duke-Williams, O. (2007) Understanding the 2001 Census interaction data: the
impact of small cell adjustment and problems of comparison with 1991, Journal of the Royal
Statistical Society Series A, 170(2): 1–21.
Chapter 11
Exploring Dimensions of School Change
During Primary Education in England
Joan Wilson
Introduction
Schools contribute to both cognitive development and the formation of non-
cognitive attributes in children. Cognitive ability concerns the knowledge procure-
ment process and involves the capacity to engage in thought and reason and to
exercise perception, judgement and awareness. Cognitive skills derive largely from
family influences and are enhanced by exposure to learning in education institutions
(see Heckman, 2000; Dearden et al., 2004). Observable measures of schooling out-
comes such as standardised tests capture both innate and acquired cognitive ability
(Postlewaite and Silverman, 2006).
That part of child cognitive development linked to education depends on the types
of schools available in the locality and the accessibility of these schools by different
households. School accessibility in an area varies according to whether institutions
are centrally funded by the government (state schools) or require the payment of fees
(independent or private schools). The focus of this chapter is on the state school sys-
tem, where entry depends on the satisfaction of school admissions criteria and over-
all intake is constrained by the pupil capacity limits faced by each school. Excess
demand for places necessitates additional joining conditions, the most significant of
which involves the proximity of a pupil’s residence to the school. At the household
level, travel costs and school quality considerations, amongst other factors, result in
preferences for entry into some local schools above others. Minimisation of travel
costs implies attendance at schools that are geographically close to the home, or that
can fit into the travel-to-work patterns of employed household members. For pupils
within households the quality of the school attended shapes cognitive learning and
affects the life course, since education outcomes are carried through to the labour
market (see Adnett and Davies, 2002). Differences in classroom teaching methods,
J. Wilson (B)
Centre for Longitudinal Studies, Institute of Education, University of London;
Centre for the Economics of Education; and Centre for Economic Performance,
London School of Economics, London, UK
e-mail: JWilson@ioe.ac.uk

212 J. Wilson
teacher quality, available subjects, the gender composition and social mix of pupils
all vary from school to school and affect overall school quality.
Non-cognitive, or life-skills – such as self-assurance, motivation, interpersonal
attributes and general emotional intellect (Carneiro et al., 2006; Heckman and
Rubinstein, 2001) – are a function of internal home experiences, schooling, and
the external neighbourhood surroundings to which a child is exposed. At home,
parent-child interactions instil the family norms, values, attitudes and behavioural
responses that impact on child non-cognitive development. Within the school these
skills relate to the nature of friendships and peer group communications that the
child engages in – both inside the classroom and when interacting with other year
groups – in addition to the overall ethos of the school. Outside of the school, resi-
dential location affects notions of opportunities that exist beyond educational years
through the amount of social capital and adult role model influences that are present
in the home surroundings, impacting on child academic aspirations and persistence
(see Glaeser, 2001). Research has revealed that life skills acquired by individuals
are advantageous not only to the individuals themselves but also to society as a
whole, since they encourage the formation of socialisation attributes and can reduce
deviant behaviour, like involvement in crime, or minimise individual exposure to
risk factors, such as unemployment or teenage motherhood (Carneiro et al., 2006).
In general the intertwined cognitive and non-cognitive facets of child progress
react to and depend upon the local provisions of services relating to schooling,
housing and other community inputs. Changes to these spatial dimensions can pro-
duce differences in the life chances of children by affecting both their cognitive
and non-cognitive development. In terms of schooling, spatial change may involve
attendance at a different school whilst remaining in the current residential location,
or may relate to a change of school occurring together with a move of home. To
date there has been limited research examining the varied aspects of moves between
schools. These seem necessary to understand if consideration is made for the impor-
tance of schools to child development.
In this chapter features of school moves excluding and involving home moves
will be described and assessed for one cohort of pupils as they progress through the
primary years of schooling that encompass Key Stage 1 (aged 6/7) to Key Stage
2 (aged 10/11) National Curriculum examinations. The empirical source of refer-
ence for this analysis is the National Pupil Database (NPD), which comprises of an
annual collection of administrative records on all state school pupils in all phases of
education throughout England since the academic year 2001/2002. The aims of this
research are: (i) to establish a means for defining ‘pure’ pupil mobility, in which
pupils change schools without moving home; (ii) to estimate the extent of school
moves made according to the definition(s) employed, and (iii) to assess the relation
between school switches and residential change, referred to throughout as ‘school-
home moves.’
The section that follows, “Literature on Mobility”, introduces evidence on the
common way of classifying and measuring pupil mobility identified in the litera-
ture and on the amount of school moves witnessed according to this method. Much
of the literature uses a composite approach in which no distinction is made between
11 Exploring Dimensions of School Change During Primary Education in England 213
school change that does and does not involve a move of home. In the section entitled
“Mobility and Government Education Policy”, recent government initiatives in the
education sector are described. These have involved an attempt to widen the scope
of education institutions that pupils can access from the current home location. At
the same time, performance indicators on the academic standards of schools have
been made publicly available, promoting the notion of parents ‘choosing’ a school
for their child to attend. All of these schemes have the potential to break the link
between school and home mobility, suggesting the need to separate out the compo-
nents to the composite measure of school change referred to in the literature.
In the “Primary Schools in England and Admissions Policies” section, a dis-
cussion of the primary school set-up in England is presented as a precursor to the
empirical focus. For pupils in LEA-governed state primary schools, the operation
of school choice is counteracted by so-called ‘over-subscription criteria’ that are
employed in the event of applications to a school exceeding available places. These
criteria can serve to reinstate the school-home connection theoretically forged apart
by school choice, implying the importance of understanding the extent to which
school changes involve home moves and how their amount compares to isolated
school shifts.
The section “Mobility in the National Pupil Database” details both the struc-
ture of the NPD and the Key Stage 1–2 cohort of primary school pupils attending
state schools in England, who make up the empirical source of reference for this
chapter. Section “Measuring Mobility in the KS1-2 Cohort” describes the means for
measuring dimensions of school mobility in this cohort and establishes the sample
size under analysis. This is followed by the “Estimating School Moves and Home
Moves” section, where such estimates are presented according to both the com-
posite measure, and to measures that divide school change only from combined
school-home moves. This empirical section highlights the loss of valuable detail on
mobility patterns that is brought about when only a composite indicator of school
moves is estimated. The richness of the NPD data source allows for assessment of
the extent to which pupils enter over-subscribed schools and how this relates to both
the kind of school move that they make and the type of state school that they enter;
evidence on these aspects is also included in the empirical part. Finally, the section
“Limitations of the Analysis” acknowledges the limitations of the analytical work
and the “Conclusions and Further Work” section concludes with a discussion of the
wider implications of this work as well as areas for further research.
Literature on Mobility
The movement of pupils between schools has been discussed in literature concerned
with the sociology of education and the management of schools as early as the
1960s (see for example, Plowden, 1967, and Douglas, 1964). In England, the first
large-scale study of the nature and causes of school change was undertaken by Dob-
son and Henthorne (1999). Their project involved the collation of general Local
214 J. Wilson
Education Authority (LEA) statistics on pupil mobility in schools, for which they
attained details from 130 out of 150 of the surveyed authorities. The authors identi-
fied four main causes of school change: international migration, internal migration,
individual movement and institutional movement. Both international and internal
migration are driven mainly by household employment factors, while individual
movement relates to changes in family circumstances which necessitate children
moving between homes. Of direct relevance to the focus of this chapter is institu-
tional movement, which involves children changing schools by choice, or transfer-
ring between differing school types, such as special and mainstream schools.
Though their analysis addresses mobility at the LEA-level, Dobson and
Henthorne (1999) also provide a general definition of mobility at the level of the
individual pupil. This is stated as “a child joining or leaving a school at a point
other than the normal age at which children start or finish their education at that
school – whether or not this involves a move of home” (p. 5, original emphasis).
Thus, pupils who switch schools at times other than when transferring from primary
to secondary school, for example, are included in the measure. This definition of
school movers is applied in many studies that assess mobility, especially those con-
cerned with the impact of school change on own-pupil educational attainment (see
for example Blane, 1985; Strand, 2002; Burgess et al., 2006). If one considers the
potential for this relationship to differ according to the form of mobility involved
then the use of such a general definition could be problematic. In this respect iso-
lated school changes, isolated home changes, or combined moves may exert varying
effects on attainment rather than having an overall clear-cut consequence.
To date there has been little research undertaken that differentiates between
school moves of various kinds, primarily due to a lack of available data that provides
detailed coverage of moves and information on their nature. Previously co-authored
work (Machin et al., 2006) has utilised the National Pupil Database (NPD) – a state-
school-level Census of pupils on roll in January of each academic year – to address
mobility patterns, where an allowance is made for differing mover types. School
moves across two waves of the Pupil-Level Annual School Census (PLASC) com-
ponent dataset to the NPD are studied, as is residential mobility behaviour asso-
ciated with in-school children, from the stance of whether there are coincident
changes of home residence occurring in line with pupil mobility. Table 11.1 details
school moves that involve a change of residence for pupils moving schools between
2001/2002 (the first wave, or year, of PLASC) and 2002/2003. Moves of school
attended are identified by changes to the code of the school recorded in the PLASC
data between the two years, and refer only to those non-compulsory school changes
rather than expected school shifts, as per Dobson and Henthorne (1999). Residential
moves reflect home postcode changes over the same period.
In the analysis of mobility patterns across all stages of education, the national
dataset used in this paper revealed that just over 900,000 school children switched
schools across the two PLASC years, equal to roughly 16% of the total of almost
5.9 million pupils sampled. Of these, just over a quarter of a million, or 4.4% of
school changes were made at non-standard time points during the academic year. It
was found that non-compulsory school changes occur most often in the transition
Table 11.1 Proportion of school movers moving home by year group and key stage (KS)
Total school Residential Proportion

Year group movers (1) movers (2) residential (3)
1–2 40,897 27,387 0.670

2–3 30,681 20,527 0.669
KS 1 average 0.669
3–4 39,606 25,188 0.636
4–5 37,007 22,948 0.620
5–6 32,577 20,984 0.644
6–7 8,808 2,365 0.269
KS 2 average 0.606
7–8 20,894 11,706 0.560
8–9 20,555 10,688 0.520
9–10 17,225 9,042 0.525
KS 3 average 0.536
10–11 8,815 4,274 0.485
School movers are pupils moving school other than at compulsory times. The remaining non-
compulsory movers between years 6–7 (when the move to secondary school occurs) reflect pupils
attending middle school who leave later than year 6. Column (1) shows total year group numbers
when both the REE school code and the home postcode are available for both academic years for
the pupil.
Source: Adapted from Machin et al. (2006, p. 264, Table 4).
from school Year 1 (aged 5/6 in 2001/02) to school Year 2 (aged 6/7 in 2002/03)
at a rate of 7.3%, and that, overall, mobility was considered to be more prevalent
in the primary school stage of education. Table 11.1 shows that more than half of
those pupils switching schools also changed residence in almost all year group tran-
sitions (except for between years 6 and 7 and years 10 and 11). Residential transfers
were higher in the primary school years, particularly in the years leading up to and
including the Key Stage 1 examinations (taken at the end of year 3, aged 6/7), at an
average of 70% for the two year groups involved (column (3)).
This evidence suggests that the school changing process is complex and analysis
of mobility patterns warrants consideration of the different forms moving can take.
In the section that follows, this point will be further highlighted in reference to
recent developments in government education policy, which have acted to affect the
nature of school moves made.
Mobility and Government Education Policy
School admissions systems can be broadly categorised into two main approaches to
schooling provision, namely community-school models and parental-choice mod-
els. In a community-based model, schools serve only local community pupils and
admission is determined purely by residential location, typically within the limits of
216 J. Wilson
a defined geographical ‘catchment’ area that comprises of pupils inhabiting homes

of close proximity to the school concerned. In the choice-based model admissions
are weighted towards parental preferences, so that parents are given more freedom
and input over the education exposure of their children. This model of education
provision incorporates a wider local area and is not restricted to place allocation in
schools nearby the home (Gibbons et al., 2006).
In recent decades, the UK government has enacted reforms pushing state school
education provision towards the choice mode as opposed to the more traditionally
featuring community system, resulting in the current existence of a hybrid edu-
cation service. A ‘quasi-market’ for the provision of education based on school
choice was first introduced through the 1988 Education Reform Act. Justifications
for this policy shift lie in the perceived merits of incentive mechanisms existing
in a competitive market place characterised by parental choice and the ability of
these devices to drive up standards in education. The application of parental choice
and school competition to the state school system has featured the formation of
market-based incentive mechanisms in school enrolment and school performance.
In terms of enrolment, school funding is linked to the number of pupils on roll at the
school. At the institution-level, league tables of performance have been supplied to
the public since 1996 (1994) for primary (secondary) schools, providing account-
ability information on the academic achievement of schools in standardised tests
relative to both nationally defined education targets and to other schools in the local
area. Taken together, these two changes mean that parents are enticed into ‘shopping
around’ for a local education supplier most satisfying the preferences and pedagogic
needs of their children and schools, in order to maximise their revenue funding, are
encouraged to actively engage in market-like competition for pupils as a result of
the policy reforms (Tiebout, 1956).
Theoretically one would expect that more transparency in the relative academic
performance of local schools and a greater parental freedom to choose amongst
a wider range of differing education providers within the same local area would
affect spatial mobility that relates to schooling, as parents attempt to take advan-
tage of opportunities for improving the learning circumstances of their children.
Under effective policy, localised changes in the school attended should be feasible
without such moves necessitating changes of home. More specifically, the empha-
sis on parental choice in education provision put forward by recent government
policy represents an attempt to sever the link between where a child lives and the
range of schools that s/he is eligible to attend, a situation imposed by the histori-
cal prevalence of education provision under the community-school model. Instead
school choice aims to forge a link between the demand for and the supply of local
education services, by offering more school alternatives conditional on pre-existing
family residential location (Gibbons and Silva, 2006a). Then it is conceivable to
suggest that such initiatives may have introduced or strengthened an element of spa-
tial activity in which pupils change schools whilst remaining in the same place of
residence, namely ‘pure’ pupil mobility.
To date empirical evidence on ‘pure’ school moves is limited, since the common
way of assessing pupil mobility in the literature has thus far failed to distinguish
between school moves only and those that involve changes of home. Given that
government policy has targeted disconnection of the home-school link, evidence on

the distinct amount of moves of each type appears important. Hence the empirical
section that follows will incorporate the redefinition and re-measurement of school
shifts, dividing them up between ‘pure’ pupil mobility and ‘school-home moves.’
One cohort of Key Stage 1 to Key Stage 2 pupils aged between 6 and 11 and engaged
in the state Primary school phase of education in England will be extracted from the
NPD for this purpose. Prior to the presentation of empirical findings, the following
section provides a brief description of the structure of Primary schooling, before
going on to address why this education stage is of particularly high relevance to the
mobility discussion.
Primary Schools in England and Admissions Policies
Primary schools in the state or ‘maintained’ schools sector in England are organ-
ised into one of four categories – Community, Foundation, Voluntary-aided (VA),
and Voluntary-controlled (VC) – where variation reflects structural differences in
governance, ownership, and pupil admissions policies, as summarised in Table
11.2. The Table shows that Community schools represent the predominant form
of state-provided primary schooling in England, catering for close to 60% of
all primary-age pupils. VA and VC schools supply education services to nearly
24% and almost 15% of primary-age pupils respectively, with Foundation schools
accounting for the remainder (just above 2%).
In terms of governance, the governing body (or board of school governors) of
a primary school is responsible for the overall management of the school – that is
it “sets the strategic direction of the school, draws up school policies, sets targets
and monitors performance” (Gibbons and Silva, 2006b, p. 8) – whilst responsibil-
ity for daily school management falls on the leadership group. In all state primary
schools, pupil funding (which is based on pupil enrolment numbers and character-
istics) is channelled from central government to each school through the LEA. VA
and VC schools are commonly attached to a ‘Foundation’, comprising of a charita-
ble (including faith) or a business organisation. Foundation schools themselves, on
the other hand, tend to operate with local organisations on a partnership basis. Over-
all ownership of school assets (the land and buildings) can belong to the LEA or to
the school governors and the principal employer of staff to the school also varies
along these dimensions.
Where pupil admissions are concerned, across all primary school types the ini-
tial coordination of admissions applications is in the domain of the LEA. In the
first instance, LEAs are required to allocate pupils to schools on the basis of stated
parental preference, as appearing in the admissions application form. For Com-
munity and VC schools so-called ‘over-subscription criteria’ are laid out by the
LEA, to be used when there are more applications to the school than places avail-
able. In Foundation and VA schools there is greater flexibility over pupil entrance
decisions, since the governing body has more freedom to set the admissions and
over-subscription rules. However, the majority representation of the ‘Foundation’,
218
Table 11.2 Characteristics of state primary schools in England
Admissions
No. of schools, authority and Assets
Type 2005/06 (% of total) Faith Governors (approximately) employer owned by
Community 8,176 (59.67%) Secular Parents >30%, Staff <30%, LEA 20%, Community LEA LEA
20%
Foundation 285 (2.08%) Mostly Secular, some Parents >30%, Staff <30%, Foundation/ Partnership Governors Foundation
C. of E. <25%, LEA <20%, Community 10% or Governors
Voluntary-aided 3,218 (23.48%) Mostly C. of E. or Catholic, Foundation >50%, Parents >30%, LEA <10%, Governors Foundation
some other faith, some Staff <30%
secular
Voluntary-controlled 2,024 (14.77%) Mostly C. of E., some other Parents >30%, Staff <30%, Foundation <25%, LEA LEA
faith, some secular LEA <20%, Community 10%
Total 13,703
The number of schools (and percentage of total) is based on the Key Stage 1 and 2 cohort data used in the empirical section and includes only those pupils
with a full set of mobility indicators. On average Community schools are the dominant institution type in England, accounting for about 65% of all Primary
schools.
Source: Adapted from Gibbons and Silva (2006b, p. 36, Table 1).
J. Wilson
as opposed to the LEA, on the board of school governors in VA schools implies that
only in these schools can admissions practices truly deviate from those applied in
LEA-run institutions (Gibbons and Silva, 2006b).
The main reason for addressing school and home moves undertaken by pupils
attending state-sector Primary schools in particular relates to both the education
sector changes introduced by the 1988 Education Reform Act and to the admissions
criteria adopted in the event of place over-subscription in Primary schools, adhered
to above. As discussed in the previous section on “Mobility and Government Educa-
tion Policy”, since 1988 government policy has favoured a quasi-market method of
education provision, in which parents are encouraged to be actively involved in the
schooling choices for their children. Theoretically, the radius of parental choice of
schools is meant to encompass a wider field than that which the traditional allocation
of pupils to localised schools would allow, and this field should be attainable from
the current residential setting. At the same time schools are encouraged to attract
a high number of pupils, since pupil quantity determines school sustainability by
being directly related to total school funding. This suggests that in the short-run
schools in high demand will be over-subscribed.
LEAs and schools are required to organise their admissions policies in line
with the current Department for Children, Schools and Families (DCSF) School
Admissions Code of Practice, which is legally binding under the School Standards
and Framework Act of 1998 (Gibbons et al., 2006). This Code also sets out the
criteria to be employed when over-subscription to a school occurs. Although poli-
cies vary by LEA, for local authority-governed Community schools, specifications
usually involve a higher chance of school entry for cases where (i) the child is of
Special Educational Needs (ii) the child has siblings attending the same school, or
(iii) the household to which the child belongs resides in the school ‘catchment area’,
an area of close geographical proximity to the school (Gibbons et al., 2006). All
over-subscription criteria are required to be both non-discriminatory and transpar-
ent, where transparency means that those LEAs in charge of school admissions make
their criteria publicly available to parents, thus serving to open up school and LEA
intake procedures. For VA schools that are not controlled by the Local Authority,
more weight is given to religious affiliation or an expression of faith by the pupil,
rather than to the satisfaction of criteria such as residential proximity to the VA
school, when there is excess demand for school places. This is in keeping with the
faith-based ethos of VA schools. Indeed it has been found that travelling distances to
VA schools generally exhibit longer area ranges than for other school types (Gibbons
et al., 2006), suggesting a more tenuous link of the school-home distance. As Table
11.2 highlights, the vast majority of Primary school pupils in England are educated
in Community schools, for whom the LEA is the admissions authority. This means
that for most pupils the over-subscription rules (i) to (iii) stated above are of greater
relevance.
Taken together, the concepts of school choice and admissions over-subscription
criteria applied to LEA-governed schools invoke a situation characterised by two
main features. Firstly, all schools that are perceived as being of good quality on
the basis of their attainment performance are likely to appear desirable to parents.
220 J. Wilson
If this holds, then this will result in an excess demand for places in such schools
and the application of the over-subscription criteria. Secondly, the publication of
these criteria gives parents a chance to influence place allocation for their child in
the over-subscribed school, if they are able to satisfy one or more of these condi-
tions of entry. The most significant condition for the concept of mobility is that of
catchment area occupancy. In respect of the geographical coverage of catchment
areas, the School Admissions Code states that “admission authorities should pro-
vide a map of the areas, and indicate how far parents within those areas have suc-
ceeded in getting places in the past, and whether that is likely to be a guide for the
future” (Department for Education and Skills, 2007, p. 53, original emphasis). Thus
parents are sufficiently informed of the catchment area space and whether living
within this space helps in ensuring place allocation to an over-subscribed school.
If pupils inhabit homes that exceed these catchment boundary limits, some house-
holds may be prepared to engage in residential mobility to within the boundary
walls in order to ensure compliance with this clause. This is likely to be a more
important aspect of spatial activity at the primary school stage in particular, because
parents will wish to maximise the quality of the school attended so as to secure
optimal future returns to schooling and at the same time they will want to min-
imise travel-to-school distances for their children in order to allay safety fears. Thus
they will have a stronger interest in relocating if this enhances the potential for
place allocation of their children in good schools nearby to the home. What this all
amounts to is a suggestion that the attempt by the parental-choice model of school-
ing provision to undo the link between residential setting and the school attended is
counteracted by the over-subscription criteria of LEA-controlled schools at the pri-
mary school stage. Hence while ‘pure’ pupil mobility may be enhanced by parental
choice provisions, over-subscription rules imply a significant ‘school-home moves’
connection. In respect of establishing some knowledge on the success or otherwise
of the quasi-market in education, it is therefore valuable to accurately define and
measure mobility patterns that relate to school moves exclusive of home moves and
combined school-home changes during the primary school years in particular. In
section that follows the empirical source of analysis to be used for this purpose is
laid out.
Mobility in the National Pupil Database
Definition and measurement of school and residential mobility will utilise a cohort
of Key Stage 1 to Key Stage 2 Primary school pupils contained in the National Pupil
Database (NPD), which itself is made up of two data sources: (i) the Pupil-Level
Annual School Census (PLASC) and (ii) National Curriculum Key Stage test scores
of attainment. In this section the origins and content of the NPD are explained, as
well as the exact structure of the cohort to be assessed.
Describing the PLASC Dataset Component of the NPD

PLASC is a unique national pupil-level administrative Census which has tradition-
ally derived information on the whole school roll in January of each academic year.
Data collection encompasses all pupils across the education spectrum of the main-
tained schools sector in England only, a situation that is enforced by the legally
binding status of the Census, in which schools are statutorily required to provide
Census information under Section 537A of the Education Act 1996 (Harland and
Stillwell, 2007). Records are supplied electronically by the school and transferred
to central government (the DCSF) via each LEA. Legal enforcement of the admin-
istrative records provision and their centralised collection reflects the use of the col-
lated statistics on pupil numbers and pupil characteristics to determine the amount
of funding to be allocated to each school (Gibbons and Telhaj, 2007). PLASC col-
lection first began in January 2001 to include pupils on roll for the academic year
2001/2002. Since 2006 (2007) a tri-annual procedure for administrative data collec-
tion was introduced into secondary (primary) schools, known as the School Census
and featuring data collection points on the third Thursday of the months of May
and September in addition to the usual (third Thursday of the month of) January
record (Harland and Stillwell, 2007). At the time of writing there are five waves
(academic years) of PLASC data available for the annual January school roll only,
yielding some 8 million pupil observations per wave, the latest being that for the
school year 2005/2006 (based on data collected in January 2006). These waves can
be linked together by means of a unique, anonymous, pupil identifier, to give a
five-year longitudinal source of PLASC information in which pupils can be tracked
as they transfer from one year group to another within the state school education
system.
PLASC contains data on individual pupil characteristics and the social back-
ground of each pupil. Important to the measurement and analysis of mobility, each
wave of PLASC includes information on the date at which the pupil entered the
school, an identifier for the school attended by the pupil, and a record of the home
postcode of the pupil, all on an anonymous basis.
Describing the Key Stage Dataset Component of the NPD

The National Curriculum was established through the 1988 Education Reform Act
and provides a standard form and content of subjects to be taught across schools
for all pupils from the age of 5–16. It was in place in all primary and secondary
schools between the academic years of 1989/90 and 1996/97. The Curriculum
divides schooling years into blocks, with each block representing a ‘Key Stage’
(KS). Curriculum comprehension is tested through national attainment examina-
tions taken at the end of each Key Stage. Formal introduction to the Key Stages
begins at the age of 5/6 (KS1) and comprises of two school years of instruction,
222 J. Wilson
leading to KS1 examinations at the age of 6/7. The KS2 phase of learning spans four
school years and final exams are sat for when the child is aged 10/11. After a further
three academic years, which include a transfer from the primary to the secondary
schooling phase (at around the age of 11), KS3 exams are taken at the age of 13/14.
These are followed by the end-of-compulsory-schooling General Certificate of Sec-
ondary Education (GCSE) exams two school years later when pupils are aged 15/16
(KS4). The publication of school average test scores in some of the Key Stages (usu-
ally KS2 and KS4 results are published in the form of ‘league tables’ appearing in
local newspapers and on the internet) enables the public and, in particular parents, to
compare the relative performance of individual local schools to other schools within
the local area and to nationally set government targets of achievement at each Key
Stage.
Information on test score outcomes for each pupil at the end of each KS is avail-
able in the NPD since the academic year 1997/98 for KS1 and KS3, 1995/96 for
KS2, and 2001/2002 for KS4. As a unique anonymous pupil identifier is included in
both PLASC and in each of the Key Stage records of the NPD this allows for one-
to-one matching of the files, such that background variables can be aligned with
attainment scores. For the analytical details presented in this chapter, the test score
information of pupils is only used to determine the exact cohort members in the
sample of interest.
Key Stage 1 to Key Stage 2 Cohort Coverage
In this chapter, one cohort of pupils contained within the NPD are studied as they
move between national KS1 and KS2 exams during state primary education in Eng-
land, where the exact form of this longitudinal sample is shown in Table 11.3.
This cohort started their KS1 phase of education in the academic year
2000/2001 at the age of 5/6 and subsequently sat for their KS1 exams in English
Reading, English Writing and Mathematics two school years later in the summer of
2002. Their KS2 phase of learning began in the school year 2002/2003 and spanned
4 academic years of instruction, leading to KS2 examinations in English, Mathemat-
ics and Science being taken in the summer of 2006. Table 11.4 details the number
of pupil-level observations for this particular cohort.
Table 11.3 The structure of the KS1 to KS2 cohort
School year group 2 3 4 5 6
Age 6/7 7/8 8/9 9/10 10/11

Key Stage 1 2
Key Stage exam year 2002 2006
PLASC academic year 2001/02 2002/03 2003/04 2004/05 2005/06
Table 11.4 Number of pupil-level observations in the KS1 to KS2 cohort
Status Number of pupils
In KS1, not KS2 (1) 23,599

In KS2, not KS1 (2) 23,908
In KS1 through KS2 (3) 552,892
Total (4) 600,399
Source: National Pupil Database (NPD).
A total of 552,892 pupils can be traced across all years of the KS1 to KS2
phases of education. PLASC records existing from the academic year 2001/2002–
2005/2006 inclusive are matched to this sample of pupils, henceforth known as
the ‘KS1-2’ cohort, using the anonymous pupil identifier available in all KS and
PLASC files. Every PLASC wave contains variables that can be used to assess
individual mobility patterns between the two Key Stages. For those cohort mem-
bers appearing in the sample in only a single KS, their mobility patterns cannot be
observed throughout the entire KS1-2 window. Observations on this group of pupils
are dropped from the sample of interest (a loss of 47,507 pupils in total, rows (1)
and (2) of Table 11.4).
Measuring Mobility in the KS1-2 Cohort
Three indicators are available to use in PLASC that allow for the measurement of
pupil and residential mobility among the KS1-2 cohort. Two of these indicators can
be applied in order to quantify school switches for individual pupils, these being the
date at which the pupil entered the school and the unique school code. The third
indicator, the home postcode of the pupil, enables evaluation of residential moves.
The exact measurement approach taken in each case is dealt with here. Beginning
with school moves, the methods are referred to as follows:
• ‘Date of school entry approach’ – this takes academic year-on-year changes to

the recorded date of entry into the school provided in the administrative data as
indicative of a school change by the pupil, so that pupil mobility = 1 if date of
school entry in year t+1 for pupil i = date of school entry in year t for pupil i
• ‘School code change approach’ – this takes changes in the recorded identifier for
the school, the school code, from one academic year to the next as indicative of
a pupil move, so that pupil mobility = 1 if school code in year t+1 for pupil i =
school code in year t for pupil i
Three key issues must be raised in respect of school mobility measured by both
of these approaches. Firstly, only those school moves taking place at non-standard
points during the Key Stage 1–2 phases are counted here. For this cohort, this means
that the following moves are not counted under pupil mobility:
224 J. Wilson
• transfers from Infant School (covering the age ranges 5–7 or 5–8) to Junior
School (age ranges 7–11 or 8–11);
• transfers from First School (age ranges 5–7 or 5–8) to Junior School (age ranges
7–11 or 8–11);
• transfers from First school (age ranges 5–8, 5–9, or 5–10) to Middle School (age
ranges 8–11, 9–12, 10–13, or 10–14).
Secondly, an intention of the analysis is to isolate ‘pure’ school moves, in which

the only dimension of the environment that is changing for the pupil is the school.
In this respect the term ‘pure’ pupil mobility is used to refer to a change of school
that does not involve any move of home in this analysis. Thirdly, in addition to pure
school shifts, pupils can also make school changes that do involve moves of home.
The unique advantage of the PLASC data source is that it allows for estimation of
the extent of combined school and home moves. In the text the term ‘school-home
moves’ will be used to address those non-standard changes of school that occur
together with a change of home.
Turning now to the actual estimation of home moves, a count of home changes
between the Key Stages can be made by comparison of PLASC records on pupil
home postcode from one academic year to the next, so that:
• Home mobility =1 if home postcode of the pupil in year t+1 for pupil i = home
postcode of the pupil in year t for pupil i
Though this forms the only method for measuring home mobility using PLASC,
it does present an accurate method when consideration is made for the geographical
proximity of postcodes: a postcode ordinarily covers at most 10 adjacent housing
units, allowing for precision in determining residential location and changes to it
(Gibbons and Telhaj, 2007).
Counting Mobility in the KS1-2 Cohort
Utilising the information contained in PLASC and the Key Stage data, Table 11.5
details the cumulative number of school moves and home moves that can be
observed and the order in which these moves appear in the data.
Tables 11.5 and 11.4 both reveal that there are a total of seven observations on
the school code, as compared with five observations on both the date of school entry
and the home postcode. This is a consequence of the Key Stage data collection
phase occurring at a different time point in the academic year relative to PLASC
data collection and the exclusion of any administrative information on the pupil
from the Key Stage files, other than the code of the school attended by the pupil
when taking their Key Stage tests. Thus pupil mobility measured according to the
school-code approach can be counted for a maximum of six times across the sample
period. This compares with a maximum count of four pupil moves using the date of
school entry approach and likewise when measuring residential mobility.
Table 11.5 Cumulative number of school and home moves in the KS1-2 cohort
Data Cumulative school moves: Cumulative
Data file collection school code school/home moves:
time date of school entry
and home postcode
PLASC January
2001/2002 2002
Key Stage Summer
1 2002 1
PLASC January
2002/2003 2003 2 1
PLASC January
2003/2004 2004 3 2
PLASC January
2004/2005 2005 4 3
PLASC January
2005/2006 2006 5 6 4
Key Stage Summer
2 2006
Source: Author’s analysis of the National Pupil Database (NPD).
In order to establish comparable measures of school moves based on the school

code method and the date of school entry method, all seven observations on the
school code cannot be fully exploited here. In fact, comparability requires that the
Key Stage information be used at most to determine the Key Stage cohort of interest
and otherwise the two school mobility estimates and home mobility will be based on
the five PLASC waves. Therefore no more than four school moves and residential
changes by the individual pupil can be observed. The under use of the school code
data may change in the future when the provision of tri-annual PLASC data (and in
particular that collection taking place in May) presents the opportunity for the closer
alignment of PLASC and Key Stage data collection points.
Accuracy of the mobility estimates requires that all pupils have an observation
on the school code, the date of school entry, and the home postcode in every PLASC
wave. That is, the amount of mobility to be derived here is to be based on a sam-
ple of pupils with a ‘full’ set of mobility variables. Without this restriction on the
sample, mobility that cannot be accounted for may occur amongst pupils with miss-
ing observations on some mobility indicators, affecting measured moves. This issue
is discussed further in the section entitled “Limitations of the Analysis”. For the
KS1-2 cohort, a ‘full’ set of observations are attained prior to any imputations to the
data and Table 11.6 below shows how this full set is altered following the imputa-
tions procedure. Detailed information on imputations and corrections to the data is
excluded from this chapter and is available from the author on request. To gauge the
extent of sample loss, the size of the KS1-2 cohort is also provided here.
Table 11.6 shows that a total of 13,505 pupil observations dropped out in the
process of merging the PLASC data to the KS1-2 cohort and in defining a sample
size containing a full set of mobility indicators based on the original contents of
226 J. Wilson
Table 11.6 Sample sizes of the KS1-2 cohort with a full set of mobility indicators
Sample type Number of pupils KS1-2 cohort sample change
(number of pupils)
KS1-2 cohort 552,892
Initial full sample* 539,387 –13,505
Imputations 6,850 +6,850
New full sample* 546,237 –6,655
Source: Author’s analysis of the National Pupil Database (NPD).
∗
The ‘full’ sample in each case is defined as that where the KS1-2 cohort member has an
observation on their school code, date of school entry, and home postcode in every PLASC
wave. The initial full sample uses the original number of observations on these indicators, prior
to any imputations or corrections. The new full sample indicates the number of additional pupil
observations that are obtained following imputations. Details on imputations and corrections
made to the data are available from the author on request. Only the imputations made (not the
corrections) affect the full sample size.
the PLASC files. Imputations that were made to both the school and home mobil-
ity indicators involved replacing a missing observation with that from the following
PLASC year when observations on an individual pupil in adjacent PLASC years to
the missing year were the same (except for in tail-end sample cases, where imputa-
tion used either the previous (if missing in 2005/06) or the following (if missing in
2001/02) year observation instead). Following imputations, pupil numbers in the full
sample rise by 6,850, giving an overall full sample size of 546,237 pupil-level obser-
vations in each of the 5 waves of PLASC and resulting in a net loss of just 6,655
pupils (1.2% of the KS1-2 cohort). It should be noted here that one might expect
these imputations to lead to a marginal underestimation of mobility amounts, if the
reason for the data to be missing relates to the pupil becoming untraceable for a
temporary period as a result of moving. The imputations process applied to the data
actually resulted in only a small sample size increase of approximately 1.3% from
the initial full amount. Also, analysis carried out on the original sample (not shown
in this chapter) revealed very similar empirical findings.
Estimating School Moves and Home Moves
It was noted in the section “Literature on Mobility” that a general definition of

school moves is given by those occurring when a child enters or exits a school
at a non-standard time point, whether or not each school change involves a move
of home (Dobson and Henthorne, 1999). However, under sections “Mobility and
Government Education Policy” and “Primary Schools in England and Admissions
Policies” it was highlighted that such a composite indicator of mobility may be inad-
equate when consideration is made for both recent advances in government educa-
tion policy and over-subscription rules applying in LEA-governed schools, both of
which suggest the importance of distinguishing between mover types. The aim of
this empirical section is to separate out and measure the amount of ‘pure’ pupil
mobility versus ‘school-home moves’ made by the KS1-2 cohort of pupils in the
NPD, as well as to discuss the implications of these results.
Composite Measures of Pupil Mobility
In Table 11.7, two composite measures of pupil mobility in the KS1-2 cohort are
estimated. These both conform to the general definition of pupil mobility referred to
above. The purpose of estimating school moves by this method is to facilitate com-
parison of the findings with those gained under definitions that allow for separation
of mobility forms.
Table 11.7 shows the number of pupils making one or more school change (that
may also involve a move of home) as they transit across the school years of the
KS1-2 phase of education. Composite measures of pupil mobility are presented that
do (columns 1 and 5) and do not (columns 3 and 7) include those compulsory school
moves that pupils have to make as part of their schooling process. These measures
are estimated utilising both the date of school entry method (columns 1–4) and
the school code approach (columns 5–8), as described in the section “Measuring
Mobility in the KS1-2 Cohort”. In comparing figures attained under the inclusion
of all school moves against those concerned with non-standard changes, it can be
seen that most compulsory school moves take place between school years 2 and 3
(when pupils are aged 6/7 and 7/8 respectively). The majority of these necessary
school moves include transfers from Infant to Junior school, yet they also comprise
of changes from First School to Junior school and First school to Middle school (as
outlined in the aforementioned section).
In order to establish a set of non-standard school changes, two techniques are
applied to the dataset so as to determine and net out required transfers. Firstly, the
postcode of each school attended by the pupil is matched to the data on an annual
basis using records on educational establishments as contained in the DCSF pro-
vided ‘Edubase’ dataset. Where the postcode of the school attended by the pupil
remains the same between one academic year and the next, but the recorded date of
school entry or the school code changes over the same years for that pupil (depend-
ing on the approach used to measure composite pupil mobility), this is taken as an
indication of a compulsory school shift. The assumption here is that if the schools
are on the same site (as is often the case with Infant and Junior schools, for exam-
ple), then the school move represents an expected change. Secondly, compulsory
school changes are removed by assessing the mobility measure itself at the school
level. Where all the pupils attending a certain school in one year move out of that
school in the following year, this is considered to be a necessary school move.
Extracting evidence on all required school changes from the measures reduces
the total amount of composite pupil mobility from 231,213 to 96,941 pupil obser-
vations under the date of school entry approach, and from 237,058 to 96,776 pupil
observations with the school code method. Out of the full sample of 546,237 pupils,
around 18% make non-standard school moves, and this is true by both the date of
school entry and the school code methods for estimating composite school change.
228
Table 11.7 Composite measures of pupil mobility across KS1 to KS2 year group transitions
All moves Non-standard All moves Non-standard

Year group (date of % of total moves (date of % of total (school % of total moves (school % of total
transitions entry) (1) sample (2) entry) (3) sample (4) code) (5) sample (6) code) (7) sample (8)
2–3 117,266 21.47 24,377 4.46 120,168 22.00 24,184 4.43

2–3; 3–4 11,042 2.02 2,861 0.52 11,225 2.05 3,059 0.56
2–3; 3–4; 4–5 1,704 0.31 515 0.09 2,026 0.37 525 0.10
2–3; 3–4; 4–5; 5–6 444 0.08 181 0.03 535 0.10 182 0.03
2–3; 3–4; 5–6 1,124 0.21 425 0.08 1,334 0.24 431 0.08
2–3; 4–5 10,148 1.86 2,286 0.42 10,176 1.86 2,274 0.42
2–3; 4–5; 5–6 1,115 0.20 357 0.07 1,206 0.22 359 0.07
2–3; 5–6 8,034 1.47 1,500 0.27 7,332 1.34 1,550 0.28
3–4 24,428 4.47 22,789 4.17 25,125 4.60 22,512 4.12
3–4; 4–5 3,579 0.66 2,489 0.46 3,844 0.70 2,488 0.46
3–4; 4–5; 5–6 490 0.09 462 0.08 551 0.10 467 0.09
3–4; 5–6 1,999 0.37 1,808 0.33 2,181 0.40 1,799 0.33
4–5 35,053 6.42 20,404 3.74 35,488 6.50 20,265 3.71
4–5; 5–6 2,191 0.40 1,856 0.34 2,436 0.45 1,867 0.34
5–6 12,596 2.31 14,631 2.68 13,431 2.46 14,814 2.71
Total 231,213 42.33 96,941 17.75 237,058 43.40 96,776 17.72
Source: Author’s analysis of the PLASC data.

The total sample size is 546,237 pupil-level observations in each of the five PLASC waves, from 2001/2002 to 2005/2006 inclusive (see Table 11.6).
J. Wilson
In fact, Table 11.7 reveals a striking similarity between school mover estimates per-
taining to the two different approaches, a situation which gives confidence in the
estimation procedure and suggests a robustness of the results to differing ways of
measuring pupil moves of school. It is worthwhile to point out at this stage that
only those school moves estimates derived under the date of school entry method
will be taken forward from here on. The reasons for this relate both to the simi-
larity of findings under the two measures and to the further analytical possibilities
provided by the date of school entry variable relative to the school code indicator.
PLASC records on the date at which the pupil entered the school specify the month
of joining and evaluation based on the use of this information will form an important
extension of future research (see Wilson, work-in-progress).
Across all year group transitions, school (and possibly home) moves are most
prevalent between school years 2 and 3, at 4.46% of the full sample, even after
correcting for compulsory transfers (column 4). Apart from this transition period,
composite mobility is also high between school years 3–4 (4.17%), 4–5 (3.74%)
and 5–6 (2.68%). In terms of the number of school moves each individual pupil
makes, the composite measure shows that it is most common for pupils to change
schools just once. As the move count rises, the number of pupils making multiple
moves falls. Between 0.27% and 0.52% of pupils make two school moves in the full
sample (two school moves account for 2.34% of the full sample overall), and the
percentage of pupils making three school moves is at most 0.09% of the full cohort
(0.32% of the full sample overall), where the later pupils move between school years
2–3; 3–4 and 4–5 (ages 6/7–9/10). Only 181 pupils (0.03%) change aspects of their
environment across every year group transition between KS1 and KS2. In terms
of school move sequences, most multiple moves involve schooling interruptions
that are made continuously, with only 0.27% of the full sample making moves that
include a gap of 2 school years (those in the transition category 2–3; 5–6).
Estimating ‘Pure’ Pupil Mobility and ‘School-Home Moves’
Table 11.8 presents the first step in analysis aimed at rooting out differences in
school move estimates according to whether or not separation of mover types is
accounted for. Here ‘pure’ pupil mobility is compared with ‘school-home moves’,
both defined in the “Measuring Mobility in the KS1-2 Cohort” section. In all cases
school moves are determined under the date of school entry approach and are exclu-
sive of compulsory school transfers of the types mentioned earlier.
A consistent pattern emerging from these results is one in which there is a dom-
inance of conjunctional school-home moves over and above pure pupil mobility
across all year group transitions featuring mobility and irrespective of the number
of school moves made. At the aggregate level, school changes that include residen-
tial change are more than 1.5 times higher among this KS1-2 cohort compared with
school moves only: 11.3% of the full sample engage in school-home moves (column
4), while 7.3% make isolated school changes (column 2). In line with the findings
of the composite pupil mobility measure shown in Table 11.7 (columns 3 and 4),
230 J. Wilson
Table 11.8 ‘Pure’ pupil mobility and ‘school-home moves’ across KS1 to KS2 year group
transitions
Year group ‘Pure’ pupil % of full ‘School-home % of full ‘New’ composite

transitions mobility (1) sample (2) moves’ (3) sample (4) school mobility (5)
2–3 11,425 2.09 15,229 2.79 26,654

2–3; 3–4 513 0.09 1,656 0.30 2,169
2–3; 3–4; 4–5 18 0.00 334 0.06 352
2–3; 3–4; 3 0.00 124 0.02 127

4–5; 5–6
2–3; 3–4; 5–6 23 0.00 240 0.04 263
2–3; 4–5 513 0.09 1,155 0.21 1,668

2–3; 4–5; 5–6 16 0.00 195 0.04 211
2–3; 5–6 223 0.04 835 0.15 1,058

3–4 9,946 1.82 15,181 2.78 25,127
3–4; 4–5 311 0.06 1,510 0.28 1,821
3–4; 4–5; 5–6 16 0.00 311 0.06 327
3–4; 5–6 256 0.05 1,088 0.20 1,344

4–5 9,509 1.74 13,047 2.39 22,556
4–5; 5–6 219 0.04 1,269 0.23 1,488
5–6 6,934 1.27 9,468 1.73 16,402
Total 39,925 7.31 61,642 11.28 101,567

School moves are measured according to the date of school entry approach and are based on non-
standard school changes only.
changers tend to make at most one school or school-home move, while pupil num-
bers are decreasing in the number of any kind of moves made. However, whereas
school moves under the composite estimates of Table 11.7 were found to be higher
during the transition between school years 2–3, here this holds true more for pupils
making school moves only. For those making school-home moves once between
school years 2–3, years 3–4 and years 4–5, their percentages of the full sample are
quite similar, at 2.79%, 2.78% and 2.39% respectively.
The ‘new’ composite measure included in Table 11.8, column (5) does not
appear, at first glance, to be comparable with that appearing earlier in Table 11.7,
column (3) (the ‘old’ composite measure), and this is actually the case. The new
composite version is the sum of pure pupil mobility (column 1) and school-home
moves (column 3). Estimation of the old composite measure is based on the general
pupil mobility definition identified in the literature. This old version does not distin-
guish between school movers of different types, whereas the new measure enforces
this distinction. This is exactly where the reason for the discrepancy between the
two estimates lies. Taking, for example, a pupil included in the mover category ‘2–3;
5–6’, under the old composite measure no details are known about whether each of
these moves are pure school moves, school-home moves, or one of each. Defining
mover types separately, it may be established that the ‘2–3’ portion of this move
represents pure pupil mobility, while the ‘5–6’ segment is a school-home move.
Then mover type separation would result in a re-classification of the school moves
of this pupil, such that their multiple move status is recorded once under the pure
pupil mobility column (1) and again under the school-move column (3), but these
counts on the pupil would be tabulated in different rows (‘2–3’ on the one hand,
versus ‘5–6’ on the other). So mover type separation enables multiple moves to be
accounted for, but a composite sum total of the different types of moves is coun-
terintuitive because there is a double-counting of pupils who make multiple moves
of differing dimensions. The difference between the total number of school movers
under the old and new composite measure reflects this. What this implies is that
there is an inaccuracy in the count of pupils categorised as making multiple moves
according to the old composite measure precisely because an individual pupil may
not always make school moves of one particular type. However, this information is
lost in the grouping together of mobile pupils as is done under the general definition
of pupil mobility, suggesting that estimation based on a separation of mover types
is much more informative in the case where a pupil changes schools more than once
and under multiple move dimensions.
Considering Entry to Over-subscribed Schools
It was explained in the section called “Primary Schools in England and Admis-
sions Policies” that LEA-governed Community and Voluntary-controlled schools
characterised by applications for places that exceed school capacity adopt over-
subscription criteria in order to rank potential entrants. One such procedure for
prioritising entry relates to catchment area occupancy, in which pupils inhabiting
homes of close geographical proximity to the school will rank higher on the school
waiting list. It was suggested that this aspect of the admissions procedure distorts
the notion of school choice, since it reduces the potential for schooling access to be
less dependent on residential location. In Table 11.8 above a distinction was made
between school movers only versus school-home movers, and it was noted that there
is more school change involving a move of home than there is pure pupil mobility in
the cohort under assessment. This finding implies that the link between the school
attended and the home setting still matters when it comes to schooling choices.
However, whether this holds true may be illustrated to some extent by the suc-
cessfulness or otherwise of pure school movers in gaining entry to over-subscribed
LEA-governed schools relative to school-home movers.
In Table 11.9, findings from a first attempt at evaluating the capacity for school
movers of the differing forms to move to over-subscribed schools are presented.
For simplicity, the analysis focuses only on those pupils moving either to over-
subscribed Community or VA schools and on those changers making only one move
of school, or school-home. Multiple movers may change the type of school that they
attend in each move made (e.g., a pupil moving schools three times may switch
232 J. Wilson
Table 11.9 Entry to over-subscribed schools

Panel A: e school movers
Year e % entering % of % % of
group pupil over- which entering which
transitions mobility subscribed coming overcoming
Community from an subscribed from an
school under- VA under-
subscribed school subscribed
school school
(1) (2) (3) (4) (5)
2–3 11,425 22.24 42.98 7.51 41.38
3– 4 9,946 24.80 48.48 8.28 51.82
4–5 9,509 23.50 60.58 8.48 61.41
5–6 6,934 20.20 63.60 8.74 61.06
Total 37,814 - - - -
Panel B: ëSchool-home movers’
(1) (2) (3) (4) (5)
Year ‘School- % entering % of % % of
group home over- which entering which
transitions moves’ subscribed coming overcoming
Community from an subscribed from an
school under- VA under-
subscribed school subscribed
school school
2–3 15,229 27.82 48.15 6.52 46.32
3–4 15,181 25.47 50.59 6.03 48.63
4–5 13,047 24.04 55.04 6.01 52.42
5–6 9,468 22.31 60.04 6.12 51.12
Total 52,925 - - - -
Movements from an under-subscribed school (columns 3 and 5) can refer to any of the
four school types (Community, Foundation, VA or VC). Pupils making one ‘pure’ school
move account for 94.71% of all pure school movers. Across all school-home movers,
pupils moving once account for 85.86% of the total (see Table 11.8).
from a Community, to a VA, back to a Community school). This complicates mat-

ters since over-subscription rules vary for more autonomously governed VA schools
compared with Community schools. In fact, as discussed in the section on “Primary
Schools in England and Admissions Policies”, VA schools are more likely to place
emphasis on factors such as religious commitment rather than catchment area satis-
faction when ranking excess pupil numbers, suggesting less of a geographical close-
ness between the school attended and the home location for pupils in VA schools.
An over-subscribed school is classified as such in the sample if the ratio of the
total number of pupils in the school to school capacity exceeds one, where annual
pupil roll and school capacity measures are obtained from the Edubase data source
as referred to earlier. These ratios are based on averaged annual pupil numbers and
school capacity figures over the five sample waves (i.e. 2001/02–2005/06 inclusive),
so as to minimise the margin for error in the over-subscribed schools indicator that
might occur were it calculated using annual ratios only.
The findings of Table 11.9 show that pupils making one school-home move
are consistently marginally more likely than pupils making one pure school move
to enter an over-subscribed Community school (column (1) of panels A and B
respectively). This suggests that the link between the school attended and residen-
tial location is important in the Primary school stage, and may be reinforced by the
catchment area criteria of filled-to-capacity LEA-governed schools. On the other
hand, some families do appear to gain access to over-subscribed Community schools
from their current place of residence, as evidenced by the pure pupil mobility fig-
ures. A further point to note is that though school home-moves to over-subscribed
Community schools feature more in this sample, the strength of the relative dif-
ference between the percentage of pupils entering an over-subscribed Community
school having made a school-home move versus a school only move falls as pupils
progress along the KS1-2 track.
Interestingly, the estimates of Table 11.9 reveal a regular pattern in each of the
year group shifts for pupils moving to over-subscribed VA schools: pupils making
one pure school move are consistently significantly more likely than those making
one school-home move to enter an over-subscribed VA school. This evidence is in
support of a more tenuous link existing between school and home proximity among
pupils in VA schools, given the tendency of these schools to adopt over-subscription
criteria related to their ethos.
It must be emphasised that these findings should be interpreted with caution.
The observation that pure school movers are gaining access to Community schools
operating above full potential could simply be because their home setting already
conforms to the catchment area clause. Indeed, for some pupils, the current home
location may be contained within the catchment area of several Primary schools,
particularly in areas with a higher density of Primary education providers, such that
catchment areas of different schools overlap the same home. In this case, one cannot
infer that the pure school move was attributable to the possibilities allowed by school
choice settings; rather it is just a reflection of extensive local education services.
At the same time, the assertion that the link between the school and the home is
reinforced by LEA-governed schools’ catchment area rules is questionable without
evidence on the reasons for the move of school and home. If the school and home
move occurred as a result of upward employment mobility, for example, then access
to a higher quality school (presumed to be so where a school is over-subscribed)
may be more an outcome of the job-related move, rather than the consequence of
a calculated move of home done so as to ensure access to a preferred school. All
of these points indicate the need for more substantive information on the nature of
school and school-home moves before any firm conclusions can be drawn about the
effectiveness or otherwise of school choice policies.
Taking into account these pitfalls, columns (3) and (5), panel A and B of Table
11.9, show an attempt to infer the extent to which school only and school-home
changes might relate to the quasi-market for schools. Estimation considers whether
the school move made to either a Community or to a VA school with an excess
demand for places involves pupils coming from a school (of any type) with spare
234 J. Wilson
capacity. Taking care again to avoid reading too much into the findings, they indi-
cate that school only movers are more likely than school-home movers to make
school changes motivated by school choice policies in the transition years 4–5 and
5–6, among those entering filled-to-capacity Community schools (column (3) com-
pared across panels A and B). The opposite is true for school only movers changing
the school attended across transition years 2–3 and 3–4. Meanwhile, for joiners
of over-subscribed VA schools who previously attended a school with spare places,
those making a school only move seem to exploit school choice opportunities earlier
on, from the transition year 3–4 (column (5) compared across panels A and B). In
line with earlier findings, this evidence reinforces the assertion that catchment area
occupancy is of less relevance as a criterion for ranking pupils when VA schools
are over-subscribed. Overall, it would appear from the estimates in Table 11.9, that
some school moves are rationalised by an interest in exploiting performance dif-
ferences between local schools, and that such school moves are able to take place
without necessitating a home change.
The findings from this empirical section show that there is much to be gained
from the detailed review of mobility patterns where movers of differing types are
separated. However, as mentioned above, there are shortcomings to the analysis
that necessitate caution when interpreting findings. The general applicability of the
estimates is also restricted to the scope of coverage of the NPD and, in the section
that follows, the nature of limitations imposed on analysis that utilises this dataset
are addressed.
Limitations of the Analysis

The focus of this work has been to look at mobility for one cohort of pupils as they
progress through state-provided schooling at the Primary stage of education, given
that recent government initiatives pertaining to the education sector involve state-
schools only. The PLASC dataset is an ideal form of secondary data to use in this
respect. However, there are limitations in the coverage of the PLASC data which
matter for the analysis of mobility patterns. One caveat is that only migration taking
place within England features in the KS1-2 cohort sample. Complete patterns of
moves among international migrants (including refugees and asylum seekers), and
among those pupils moving from elsewhere within the UK who enter a school in
England for a certain length of time, cannot be established. At best only one KS
test outcome may exist for such pupils. For schools in the cities and metropolitan
areas of England this type of pupil entry and exit will make up a large proportion of
their school joiner-leaver activity. Then all mobility measures will be understated by
that amount of movement that reflects cross-country migration, and this shortfall of
the data can produce a non-trivial flattening of regional variation. Secondly, where
households opt out of state provided education and buy into the schooling provisions
of the private fee-charging sector and for those moving in the opposite direction, res-
idential and school changes assessed using PLASC will be understated by the omis-
sion of independent school pupils in the data. Thirdly, children who are schooled at
home and who may or may not be instructed in line with the requirements of the
National Curriculum will not feature in the data since they will not be enrolled in
a publicly-provided learning institution. In the dataset all of these exclusions from
PLASC – international migrants, independent school pupils, and home-tutored chil-
dren – are likely to form those observations where there is attrition at some point in
the sample presence, as detailed in Table 11.4 and Table 11.6.
In terms of the methods for estimating the amount of mobility presented in this
work, in all approaches there is no means for assessing multiple within academic
year pupil moves, since the administrative data on the school roll used here is col-
lected only once per year. Provision of tri-annual PLASC data represents a major
step forward for future projects concerning mobility. However, the fact that PLASC
is able to provide a longitudinal panel of observations on the same pupil as they
move through the schooling years does mean that multiple year-on-year moves of
both school and home can be considered. It should be emphasised that this kind of
cohort analysis of moving activity in general, and also of mobility of different types,
has not been feasible on such a large scale prior to the introduction of PLASC: thus
the dataset acts a significant resource for researchers.
Finally, in the sample used here there is the potential for some moves to be taking
place towards the end of the KS2 phase of education that cannot be observed in the
data, leading to the underestimation of total moves over the entire KS1-2 phase. In
particular mobility taking place between January 2006 (when PLASC 2005/2006 is
collected) and the summer of 2006 (when KS2 tests are taken) is unaccounted for in
the data (see Table 11.5). However, this is a very small window of missing data of
at most 6 months, assuming that pupils take their KS2 tests in July at the very latest.
Of greater concern is the likelihood that much moving activity actually takes place
outside of the entire KS1-2 cohort sample frame. Indeed, the evidence presented in
Table 11.1 revealed combined school-home mobility to be at its highest during the
school year 1–2 transition. Then, if this holds true in the current sample, such moves
cannot presently be observed. This means that the analysis will omit important early
years moving behaviour that is likely to be of direct relevance to the moving patterns
witnessed in this sample. As the longitudinal nature of the PLASC dataset widens
in the future, this will increase the breadth of detail on mobility for older KS1-2
cohorts than that sampled here and for other cohorts of pupils who can be tracked
across KS test years.
Conclusions and Further Work

Changes to government education policy since the late 1980s and the operation
of state primary school admission systems have both had the potential to alter the
nature of school moves that pupils may undertake in differing ways. On the one
hand school choice has aimed to expand the capability of pupils to access a wider
amount of schools from the same local area. On the other hand school popularity has
required some form of admissions procedure to be put in place to allow for the rank-
ing of pupil entry to schools operating above full capacity. It has been suggested here
that a catchment area occupancy rule has served to maintain the link between where
236 J. Wilson
a child lives and the range of schools they can attend, limiting the market place
for schooling that education reforms aimed to develop. Whilst the empirical analy-
sis undertaken here does not provide concrete evidence on whether this admissions
rule has impacted on the link between the home and the school, evidence from other
research proposes this to be the case. In particular, recent analysis of house prices
and school quality in the UK has shown that parents of primary school-age children
are relocating to a residence within the catchment area of a high-performing local
primary school in an attempt to secure a place for their child in the establishment,
often paying significant house price premia in order to do so. Estimates suggest that
a 10% point rise in the number of pupils achieving Level 4 national target grades
in their end of KS2 tests adds around a 3% property price premium to houses in
London and the surrounding Metropolitan areas (Gibbons and Machin, 2006).
Evidence arising from this work does partially indicate that a portion of moves
across state primary schools may reflect the exploitation of differentials in school
quality. One direction of research currently underway involves studying the exact
relation between pupil-level moves of school and school quality (as measured by
performance tables) for the KS1-2 cohort addressed here (see Wilson, work-in-
progress). The intentions of this undertaking are to determine the extent to which
pupils who change only their school and pupils for whom this is combined with
a home move are able to make education quality gains and to ascertain how these
angles relate to the background characteristics of moving pupils. This process of
analysis will serve to enrich and potentially reinforce the preliminary findings on
the general effectiveness of school choice policies presented here.
One dimension of mobility that has received scant attention so far, largely due
to the under-provision of data that can be directly applied to the issue, is that of the
extent of ‘home mobility’ that pupils engage in. A natural extension of the work
presented here, where the differing types of school moves have been assessed, is to
apply the same concepts to residential change. Then where this analysis has focused
on ‘pure’ pupil mobility and ‘school-home moves’, evidence can be extended to
provide details on the amount of ‘pure’ home mobility and ‘home-school moves’
that pupils undergo. Gaining some understanding of the extent of home change that
occurs amongst school age pupils is important to establish, given the tendency for
research to focus more heavily on the notion of school moves of any kind affecting
education progression (see Wilson, work-in-progress, for further details).:
Acknowledgments I would like to thank Kirstine Hansen, Heather Joshi, Ruth Lupton, Olmo Silva
and Anna Vignoles for their very useful comments and suggestions on a draft version of this work.
I would also like to thank John Stillwell for his editorial contributions and the DCSF for granting
me access to the NPD data. All errors and omissions are my own
References
Adnett, N. and Davies, P. (2002). Markets for Schooling: An Economic Analysis. London:
Routledge Research in Education.
Blane, D. (1985). A Longitudinal Study of Children’s School Mobility and Attainment in Mathe-
matics. Educational Studies in Mathematics, 16(2): 127–142.
Burgess, S., Goldstein, H. and McConnell, B. (2006). Modelling the Impact of Pupil Mobil-
ity on School Differences in Educational Achievement. The Centre for Market and Pub-
lic Organisation (CMPO) Working Paper Series No. 156, University of Bristol, Bristol.
www.bris.ac.uk/depts/CMPO/workingpapers/wp156.pdf. Accessed 29 April 2008.
Carneiro, P., Crawford, C. and Goodman, A. (2006). Which Skills Matter? Centre for the Eco-
nomics of Education Discussion Paper No. 59, London School of Economics and Political
Science. cee.lse.ac.uk/cee%20dps/ceedp59.pdf. Accessed 29 April 2008.
Dearden, L., McGranaham, L. and Sianesi, B. (2004). The Role of Credit Constraints in Edu-
cational Choices: Evidence from NCDS and BCS70. Centre for the Economics of Edu-
cation Discussion Paper No. 48, London School of Economics and Political Science.
cee.lse.ac.uk/cee%20dps/ceedp48.pdf. Accessed 29 April 2008.
Department for Education and Skills (2007). School Admissions Code, The Stationery Office.
www.dfes.gov.uk/sacode/docs/DfES%20Schools%20text%20final.pdf. Accessed 29 April
2008.
Dobson, J. and Henthorne, K. (1999). Pupil Mobility in Schools: Interim Report. Migration Unit
Research Report No. 168, Department of Geography, University College London, London.
Douglas, J. W. B. (1964). The Home and the School. MacGibbon and Kee, London.
Gibbons, S. and Machin, S. (2006). Paying for Primary Schools: Admission Constraints, School
Popularity or Congestion? The Economic Journal, 116, Number 510: C77–C92.
Gibbons, S., Machin, S. and Silva, O. (2006). Competition, Choice and Pupil Achievement. Journal
of the European Economic Association, 6(4): 912–947.
Gibbons, S. and Silva, O (2006a). Competition and Accessibility in School Markets: Empirical
Analysis Using Boundary Discontinuities. Chapter 7, Vol. 14 in Gronberg, T. J. and Jansen,
D. W. (eds) Improving School Accountability – Check-Ups or Choice. Advances in Applied
Microeconomics series, Elsevier, pp. 157–184.
Gibbons, S., and Silva, O. (2006b). Faith Primary Schools: Better Schools or Better Pupils? Centre
for the Economics of Education Discussion Paper No. 72, London School of Economics and
Political Science. cee.lse.ac.uk/cee%20dps/ceedp72.pdf. Accessed 29 April 2008.
Gibbons, S. and Telhaj, S. (2007). Mobility and School Disruption. Centre for the Economics
of Education Discussion Paper No. 83, London School of Economics and Political Science.
cee.lse.ac.uk/cee%20dps/ceedp83.pdf. Accessed 29 April 2008.
Glaeser, E. (2001). The Formation of Social Capital. Isuma, 2(1): 34–40. www.isuma.net/v02n01/
glaeser/glaeser e.pdf. Accessed 29 April 2008.
Harland, K. and Stillwell, J. (2007). Commuting to Schools in Leeds: How useful is the PLASC?
Working paper 07/02, School of Geography, University of Leeds, Leeds.
Heckman, J. J. and Rubinstein, Y. (2001). The Importance of Noncognitive Skills: Lessons from
the GED Testing Program. The American Economic Review, 91(2): 145–149.
Heckman, J. (2000). Policies to Foster Human Capital. Research in Economics, 54(1): 3–56.
Machin, S., Telhaj, S. and Wilson, J. (2006). The Mobility of English School Children. Fiscal
Studies, 27(3): 253–280.
Plowden, B. (1967). Children and Their Primary Schools. HMSO, London.
Postlewaite, A. and Silverman, D. (2006). Non-cognitive Skills, Social Success, and Labor Mar-
ket Outcomes. Working paper, University of Pennsylvania. www.ssc.upenn.edu/∼apostlew/
paper/pdf/ns.pdf. Accessed 29 April 2008.
Strand, S. (2002). Pupil Mobility, Attainment and Progress During Key Stage 1: A Study in Cau-
tious Interpretation. British Education Research Journal, 28(1): 63–78.
Tiebout, C. M. (1956). A Pure Theory of Local Expenditures. The Journal of Political Economy,
64(5): 416–424.
Wilson, J. (work-in-progress) Geographical mobility, pupil mobility and school choice in England,
PhD, Institute of Education, University of London, London.
Index
A timing, 24, 25, 31

Abortion, 76, 83, 85 Bracken School Readiness, 72
Active ageing, 12, 172, 187 Brain drain, 13
Active childcare, 179–180 British Ability Scales (BAS) Naming
Administrative data, 15, 221, 223, 235 Vocabulary, 72, 73, 74
Adolescents, 99, 109, 154, 155, 166, 167 British Cohort Study (BCS), 16, 27, 28, 63, 64,
Adopted children, 28 65, 67, 70
Adulthood, 27, 154, 157 British Household Panel Survey (BHPS), 9,
Age at birth, 47, 63 14, 16, 34, 42, 44, 45, 107, 108, 109,
Ageing, 1, 8, 10, 12, 16, 127, 128, 130, 172, 110, 112–121, 192, 193, 201, 202, 203,
180, 187 204, 205, 206, 207, 208, 209, 210
Age-specific fertility rate, 3, 48 British Social Attitudes (BSA), 107, 108, 110,
Alcohol consumption, 85 111, 180, 186
Anxiety, 86, 158
Assisted conception, 145–146, 174–175, 176 C
Assortative mating, 44, 83 Canada, 129, 186
Asylum seeker, 15, 234 Care, 1–16, 44, 60, 66, 70, 72, 74, 77, 84, 85,
Attitudinal data, 106, 107 129, 130, 139, 144, 145, 171–187, 233
Attrition bias, 28 Career, 16, 24, 45, 54, 55, 66, 76, 77, 83
Australia, 69, 120 Caring, 12, 13, 24, 172, 174, 176, 177, 178,
Austria, 132, 174, 178, 180, 180, 185
185, 186 Census, 12, 13, 14, 15, 16, 86, 131, 132, 133,
138, 139, 141, 192, 193, 194, 195, 196,
B 197, 198, 199, 200, 201, 202, 208, 209,
BCS70, 16, 65, 66, 67, 70, 85, 86, 87, 88, 89, 214, 220, 221
90, 91, 92, 93, 94, 95, 96, 97, 98, 99, Census Origin-Destination Statistics, 16
100, 101 Census Special Migration Statistics (SMS),
Behavioural disorders, 158 192, 193, 194, 195, 197, 198, 201
Behavioural problems, 156, 158, 160, 161, Centre for Epidemiological Studies eight-item
162, 163, 164, 165, 166, 167 scale of depressive symptoms (CES-D),
Belgium, 11, 132, 174, 178, 180, 185 135, 136
Bias, 6, 28, 47, 48, 118, 135, 156, 179 Childbearing, 2, 5, 7, 23–38, 41–57, 61, 62, 63,
Biological parent, 151, 152, 153, 166 65, 66, 67, 75, 82, 83–84, 99, 100, 105,
Birth 128, 172, 175
cohort, 4, 16, 27, 42, 46, 51, 53, 55, 60, Childcare, 4, 5, 6, 12, 33, 43, 44, 46, 54, 56,
61, 62, 63, 64, 67, 70, 76, 86, 108, 109, 57, 60, 69–75, 76, 77, 78, 171, 172,
154, 180–181, 184, 187 173–178, 179, 180, 181, 182, 183, 184,
interval, 23–24 185, 186, 187
spacing, 24 Child-dependency ratios, 2
239
240 Index
Child development, 10, 16, 27, 63, 71, 72, 73, Death, 2, 47, 85, 129, 131, 132, 133, 139,
77, 78, 86, 154, 155, 212 142, 173
Childlessness, 4, 5, 23–38, 42, 49, 50, 65–66, Deinstitutionalisation of marriage, 107
89, 97 Delayed childbearing, 23–38, 82, 172, 175
Child migration, 15 Démariage, 107
Child outcomes, 5, 6, 56, 59–78, 179 Denmark, 129, 132, 136, 144, 174, 178, 180,
Children, 1, 2, 3, 6, 7, 9, 10, 12, 15, 23–29, 181, 185
33, 34, 35, 41, 42, 43, 44, 46, 47, 48, Department for Education and Skills (DfES),
53–57, 59, 60, 61, 62, 66, 67, 69–78, 15, 70, 71, 220
82, 83, 84, 92, 94, 97, 98, 100, 108, Dependency ratios, 2, 11, 12
111, 115, 128, 129, 130, 134, 138, Dependent children, 10, 12, 33, 59, 67, 173,
151, 152, 153, 154, 155, 157, 158, 174, 204, 207
159, 160, 161, 162, 164, 165–166, Depression, 74, 129, 132, 133, 134–136, 137,
167, 171–187, 191, 193, 195, 202, 153, 156, 158, 166
204, 207, 211, 212, 214, 216, 219, Diaspora, 13
220, 234 Direct observation, 179
Cognitive function, 71, 75, 78, 83, 84, 130, Disability, 12, 129, 145
211, 212 Disestablishment of marriage, 107
Cognitive outcomes, 60, 61, 71, 72, 77, 152 Divorce, 2, 9, 12, 28, 37, 55, 84, 85, 105, 114,
Cognitive skills, 211 132, 151, 152, 153, 167, 172, 175, 192
Cohabitation, 2, 7, 8, 9, 26, 33, 34, 35, 37, Divorce rate, 12, 151, 167, 172
41, 54, 63, 105, 106, 107, 108, 109, Domestic tension, 156, 158, 160, 161
110, 111, 112, 113, 114–119, 120, 121, Domestic violence, 9
151, 192 Domiciliary care, 12–13
Cohabiting couples, 105, 111, 112, 114, 118, Drug use, 85, 152
119–120 Dual-earner-dual-carer, 177
Cohort analysis, 26, 52, 235 Dual-earner family, 177
Cohort fertility, 48 Dual stepfamilies, 158, 159, 161, 162, 163,
Communal establishment, 129, 130, 131, 196 164, 165, 166, 167
Community schools, 215, 216, 217, 218, 219,
232, 233, 234
Commuting, 1, 13, 14, 15, 178, 191, 194 E
Companionship, 7, 108, 115, 129 Early parenthood, 5, 6, 81–101
Completed family size, 27, 43 Education, 2, 3, 4, 5, 6, 8, 15, 24, 27, 28,
Connectivity index, 197, 198, 199 30, 31–32, 34, 37, 41–57, 61, 62, 64,
Contraception, 24, 83 65, 66, 68, 69, 70, 71, 72, 75, 76, 77,
Co-residential, 9, 10, 28, 113, 119–120, 83–84, 89, 91, 93, 96, 98, 99, 135, 152,
121, 173 156, 158, 160, 161, 164, 165, 211–236
Culture, 7, 10, 12, 131, 144, 176, 186 Educational attainment, 24, 28, 30, 31–32, 37,
Czech Republic, 11, 132, 180 52, 55, 56, 61, 77, 96, 152, 214
Educational qualification, 4, 31, 32, 37, 42, 43,
D 137, 140, 158
Daily mobility, 13 Education policy, 213, 215–217, 219, 226, 235
Data, 2, 6, 7, 9, 10, 12, 13, 14, 15, 16, 25, 26, Education Reform Act, 216, 219, 221
27–28, 32, 35, 42, 44, 45, 46–48, 53, Emotional well-being, 128, 132, 134, 135, 144
56, 60, 63, 65, 66, 70, 72, 84, 85, 86–87, Employment status, 156, 160, 161, 183
89, 92, 95, 96, 97, 106, 107, 108, 110, Engagement, 5, 97, 120
112–114, 119, 120, 128, 130–133, 135, England, 10, 12, 15, 25, 26, 47, 62, 63, 64,
138, 145, 154, 155, 157, 162, 164, 165, 70, 105, 120, 128, 129, 130, 131, 132,
166, 167, 172, 175, 176, 179, 180, 133–136, 137, 138–143, 144, 145, 193,
181, 183, 184, 187, 192, 193–196, 198, 211–234
199, 201, 202, 204, 208, 209, 212, English Longitudinal Study of Ageing (ELSA),
213, 214, 218, 220–228, 230, 232, 234, 16, 130, 131, 132, 133, 134, 135, 136,
235, 236 137, 138, 139, 140, 141, 143, 145, 180
Index 241
Equal Pay Act, 6 G

Ethnicity, 7, 15, 16, 74, 85 Gender, 4, 7, 12, 15, 25, 31, 32, 37, 86, 88, 89,
Ethnic minority, 66, 85 97, 100, 112, 133, 134, 136, 138, 139,
Europe, 1, 3, 8, 10, 12, 16, 23, 26, 33, 43, 143, 160, 172, 175, 177, 185, 187, 195,
61, 105, 107, 127, 128, 130, 131, 132, 203, 204, 212
133–134, 135, 136, 138, 143–144, 145, equality, 32, 177
171, 172, 175, 177, 178, 180, 181, 184, equity, 25
185, 186 Generational contract, 11, 12
European Community Household Panel, Germany, 11, 132, 174, 177, 178, 180, 185,
16, 180 186
European Council, 174, 177 Globalisation, 8
European Social Survey (ESS), 16, 130, 131, Grandchild(ren), 12, 171–187
132, 133, 134, 136, 138, 143, 181, Grandfather, 172, 175, 179, 185, 187
184, 185 Grandmother, 172, 175, 179, 183, 185, 187
European Union, 33, 174, 176, 177 Grandparent, 12, 69, 70, 72, 74, 75, 77, 82,
Eviction, 207 171–187
Grandparenthood, 12, 175
F Great Britain, 13, 26, 27, 120, 136, 185,
Family, 1–10, 12, 16, 24, 25, 27, 28, 33, 41, 193, 195
42, 43, 44, 46, 47, 49, 50, 51, 53, 55, Greece, 132, 174, 178, 180, 185
56, 60, 61, 62, 66, 67, 70, 74, 78, 84, Growing up in Scotland (GUS), 16, 181, 182,
85, 86, 94, 97, 105, 112, 128, 129, 131, 183, 184
144, 145, 151–167, 172, 173, 176, 177,
178, 179, 180, 183, 184, 187, 191, 199, H
204, 207, 208, 211, 212, 214, 216 Happiness, 7, 9, 131, 132, 133, 136–138, 145
Family Expenditure Survey (FES), 16, 46, 47, Health, 9, 10, 12, 14, 27, 47, 56, 61, 71, 72,
48, 49, 50, 51, 52, 53, 54, 56 84, 127–146, 151–168, 172, 173, 180,
Family Resources Survey (FRS), 16, 46, 47, 193, 207
48, 49, 50, 51, 52, 53, 54, 56 Healthcare, 84
Family type, 156, 157, 159, 161, 162–164, Health selection, 128, 129, 130, 132, 141, 142,
165, 166, 167 143, 145, 160
Fatherhood, 4, 6, 83, 84, 85–86, 87, 88, 89, Health Survey for England (HSE), 131
92–94, 95, 96, 97, 99, 100 Higher education, 2, 5, 6, 43, 44, 45, 64, 84
Female employment, 24, 45, 56, 106, 178 Higher order marriages, 8
Fertility, 1–16, 23, 24, 25, 26, 27, 28, 31, 32, Home, 12, 15, 28, 33, 34, 47, 48, 59, 74, 75,
36, 37, 38, 42, 43, 44, 46, 47, 48, 49, 50, 77, 84, 129, 130, 144, 152, 153, 156,
52, 59, 62, 64, 65, 76, 81, 82, 87, 97, 157, 192, 193, 195, 211, 212, 213, 214,
151, 172 215, 216, 217, 219, 220, 221, 223, 224,
Financial hardship, 157, 158, 160, 161 225, 226–234, 235, 236
Financial independence, 127 Home Behaviour Scale (HBS), 156, 157, 158,
Finland, 11, 132, 144, 174, 178, 161, 162, 163, 164
185, 186 Homogamy, 83
First demographic transition, 2 Household
First families, 152, 153, 154, 156, 157, 158, composition, 151
159, 161, 162, 163, 164, 165, 166, 167 income, 55, 56, 61, 74
Flexible working, 60 structure, 2, 193
Formal care, 12, 60, 70, 171, 172, 173, 174, survey, 46, 105, 179, 180
175, 176, 181, 187 Housework, 179
Formal childcare, 5, 44, 46, 56, 57, 60, 69, 70, Housing, 34, 74, 83, 89, 95, 99–100, 101, 137,
72, 77, 78, 172, 173–176, 183, 186, 187 140, 193, 212, 224
Foster care, 85 Housing tenure, 89, 99–100, 101,
Foundation schools, 217 137, 140
France, 11, 43, 131, 174, 178, 180, 181, 185 Hungary, 132, 174–175
242 Index
I Lone parent, 14, 60, 152, 156, 157, 158, 159,

Ill-health, 9, 12 161, 162, 164, 165, 166, 167, 182,
Immigration, 11, 13, 66 204, 207
Income, 24, 43, 44, 45, 46, 55, 56, 61, 74, 84, Lone parent family, 159, 161
130, 154, 164, 183, 193 Longitudinal analyses, 131, 167
Individual attitudes, 111–112 Longitudinal data, 27, 128, 145, 154
Individualisation, 8, 37, 127 Luxembourg, 131, 174
Informal care, 60, 70, 171, 172, 174, 175, 176,
181, 187 M
Informal carer, 171, 172, 173 Malaise Inventory Scale (MIS), 156, 157
Inherited wealth, 11 Male fertility, 6
Intentions, 85, 90, 91, 97, 98, 111, 112, 113, Marital quality, 152, 153, 154, 167
116, 117, 119, 120, 205, 206, 224, 236 Marriage, 2, 7, 8, 25, 26, 33, 34, 35, 36, 37, 41,
Internal migration, 191–210, 214 54, 61, 63, 64, 75, 105, 106, 107, 108,
International migration, 1, 13, 214 109, 110–120, 151, 152, 153, 167, 192
Ireland, 11, 15, 131, 174, 178, 180, 181, 185, Marriage rate, 2, 105
193, 194 Maternal employment, 5, 55, 56, 57, 59, 60,
61, 66–69, 71, 72, 76, 77, 78
J Maternal grandparents, 175
Japan, 129 Maternity leave, 4, 6, 54, 60, 68, 69, 76
Men, 4, 6, 7, 12, 24, 25, 26, 28, 29, 30, 31, 32,
K 33, 34, 35, 36, 37, 41, 45, 65, 83, 87,
Key Stage, 15, 212, 213, 215, 217, 218, 220, 100, 109, 113, 115, 116, 117, 118, 119,
221–223, 224, 225 120, 133, 134, 135, 136, 137, 138, 139,
140, 141, 142, 143, 144, 145, 152, 153,
L 160, 171, 172, 177, 185, 186
Labour Mental health, 9, 10, 61–62, 84, 128, 136, 138,
force, 2, 12, 24, 32, 45, 76, 105, 106 151–168
market, 3, 4, 5, 6, 12, 13, 24, 25, 27, 43, 60, Migrant worker, 13, 15
61, 65, 66, 67, 68, 69, 76, 83, 84, 98, Migration, 1, 11, 13, 14, 15, 66, 191–210,
99, 164, 171, 175, 176, 177, 178, 186, 214, 234
187, 211 Migration connectivity, 197–201, 208
Leisure, 172, 187 Millennium Cohort Study (MCS), 16, 59, 63,
Life course, 1, 8, 27, 65, 77, 83, 153, 155, 158, 64, 66, 67, 68, 69, 70, 72, 76, 77, 85,
167, 175, 187, 211 181, 182, 183
Lifecycle, 10 Mobility, 1–16, 90, 172, 191, 192, 193, 194,
Life expectancy, 10, 12, 127, 128, 134, 172 195, 197, 201, 202, 203, 208, 212,
Lifeways Cross-Generation Cohort Study, 181 213–217, 218, 219, 220, 221, 223–226,
Limiting long-term illness (LLTI), 132, 135, 227, 228, 229–231, 232, 233, 234,
137, 141, 142, 145 235, 236
Live-in union, 113 Mortality, 1, 11, 47, 128, 129, 134, 138–143,
Living alone, 10, 26, 34, 127, 128, 129, 130, 144, 152, 172
132, 133, 135, 136, 137, 139, 140, 141, Motherhood, 2, 3, 4, 5, 6, 24, 43, 45, 51, 52, 53,
144, 145, 146 54, 55, 56, 59–78, 81, 82, 83, 84, 85,
Living-apart-together (LAT), 8, 120 86, 87, 88, 92–94, 95, 96, 97, 98, 99,
Living arrangements, 1–16, 25, 34, 105, 106, 100, 212
107, 120, 121, 127–146 Multi-generational households, 11, 127, 128
Local Education Authority (LEA), 213–214, Multinational Time Use Study, 180, 186
217, 218, 219, 220, 221, 226, 231, 233
Logistic regression, 88, 133, 135, 136, 137, N
139, 140, 142, 144, 158 National Childcare Strategy, 77
Loneliness, 9, 129, 130, 132, 133, National Child Development Study (NCDS),
134–136, 137 10, 16, 27, 28, 63, 64, 65, 66, 67, 70,
Lone mother, 41, 54, 55, 56, 69, 76, 152, 174 71, 76, 77, 85, 86, 87, 88, 89, 90, 92,
Index 243
93, 94, 95, 96, 97, 98, 99, 100, 154, Postponement of marriage, 25–27, 33–35, 41,
155, 156, 159, 160, 164, 165 54, 105, 120
National Pupil Database (NPD), 15, 212, 213, Poverty, 5, 13, 60, 67, 69, 100, 101, 152, 172
214, 217, 220–223, 225, 226, 227, 234 Primary school, 71, 183, 213, 215, 217–220,
Neighbourhood, 99, 183, 212 221, 226, 231, 233, 235, 236
Netherlands, 132, 173, 174, 178, 180, 181, Prison, 130, 173
185, 186 Privacy, 127, 129
Never-married, 9, 28, 34, 35, 36, 55, 114, 116, Productive generation, 12
117, 120, 132, 139 Property ownership, 8
New Deal for Lone Parents, 60 Psychological health, 129, 153
Non-cognitive skills, 212 Psychological well-being, 151
Non-marital birth, 25, 63 Public policy, 128, 144
Normative attitudes, 106–111 Pupil Level Annual School Census (PLASC),
Northern Ireland, 15, 193, 194 15, 16, 214, 220–221, 222, 223, 224,
225, 226, 228, 229, 230, 232, 234, 235
O
Office for National Statistics (ONS) Q
Longitudinal Study (LS), 16, 26, 27, Qualitative research, 9, 26, 84, 121
48, 62, 63, 64, 105, 130, 131, 132, 133, Quality of life, 129
134, 138, 139, 140, 141, 142, 143, 144
Old age dependency ratio, 2, 11 R
Older people, 1, 10, 11, 12, 127, 128, 129, 130, Reasons for moving, 192, 201–208
138, 143, 144, 145, 146, 171, 172 Recall error, 6
Opinion poll, 106 Registration data, 15, 48
Over-subscribed school, 213, 220, 231–234 Relationship
Own child method, 46, 47, 48 breakdown, 207
expectations, 9, 111, 112, 114, 116, 117,
P 118, 119
PACS, 8 formation, 8, 207
Paid care, 44, 54 outcomes, 112, 114, 117, 118, 119
Panel data, 110, 167, 204 Religion, 7, 28
Parental failure, 173 Remarriage, 152, 153, 167
Parenthood, 2, 3, 4, 5, 6, 9, 10, 12, 24, 25, 26, Remittance, 13
37, 66, 67, 81–101, 112, 115, 120, 152, Re-partnering, 175
154, 167, 175, 176 Replacement level fertility, 2
Parenthood status, 37, 115 Reproduction, 6, 25, 106
Parenting style, 155 Residential location, 191, 212, 215, 216, 224,
Partnership dissolution, 34 231, 233
Part-time employment, 52, 54, 178 Residential mobility, 1, 13, 192, 193, 214, 220,
Passive childcare, 180 223, 224
Paternal grandparents, 175 Retirement, 175, 180, 187, 204
Paternity leave, 4, 60 Retrospective survey question, 179
Peer group, 100, 212 Risk
Pension, 10 avoidance, 8
Philoprogeneity, 93, 100 behaviours, 9
Pivot generation, 11 identification, 8
Policy, 1, 3, 5, 6, 13, 16, 24, 65, 68, 77–78, 96, Rutter score, 86, 92
107, 128, 144, 145, 172, 177, 187, 213,
215–220, 226, 235 S
Population at Risk (PAR), 192 Same sex unions, 8
Portugal, 132, 173, 174, 175, 178, 185 Sample attrition, 194, 201
Post-marital cohabitation, 8 Satisfaction, 130, 131, 153, 211, 219, 232
Post-materialist values, 23 Scandinavia, 131, 145
Postponement of fertility, 23–24, 28 School
244 Index
admissions, 211, 215–216, 219, 220, 235 Stepparent, 10, 151–168

children, 44, 53, 54, 55, 56, 193, 214 Stepparenting, 8, 10, 151–168
choice, 213, 216, 219, 231, 233, Stepson, 155–160
234, 235, 236 Stress, 9, 129, 153, 154, 155, 164, 167
pupil, 14, 212, 213, 219, 220, 234–235 Student, 15, 24, 43, 45, 47, 192, 193, 195, 196
quality, 211, 212, 236 Substance abuse, 173
Scotland, 15, 16, 180, 181, 182, 184, 193, 194, Sure Start, 60, 78
198, 199 Surrogate parenting, 173, 175
Secondary school, 49, 214, 215, 216, 221 Survey data, 26, 29, 30, 32, 35, 36, 106, 157,
Second demographic transition, 2, 23, 28, 59, 162, 163, 175
81, 105 Survey of the European Community Household
Secularisation, 8 Panel (ECHP), 16, 174, 180
Self-rated health, 132, 133, 134, 138, 139, 140, Survey of Health, Ageing and Retirement in
141, 142, 143–144, 145 Europe (SHARE), 145, 180, 184, 185
Self reports, 179 Survival analysis, 47
Separated, 28, 114, 132, 151, 152, 234 Sweden, 11, 43, 107, 128, 130, 132, 144, 178,
Separation, 8, 25, 37, 55, 66, 85, 106, 112, 152, 180, 185, 186
192, 227, 229, 231
Sexual abuse, 84 T
Sexual behaviour, 93, 94 Teenage fatherhood, 92–94, 96, 97, 99, 100
Single person household, 10, 198, 199, 200 Teenage motherhood, 3, 5, 60, 61, 78, 82, 85,
92–94, 96, 97, 98, 100, 212
Singleton, 13, 192, 195
Teenage parenthood, 5, 82, 89
Social care, 12, 145
Tenure, 6, 84, 89, 90, 91, 92, 94, 95, 99–100,
Social Chapter, 177
101, 137, 140, 141, 193
Social class, 4, 6, 28, 29–31, 37, 84, 89, 90, 91,
Tertiary education, 31, 32, 37, 65, 76
99–100, 101, 140, 141, 156, 158, 160,
Time allocation, 179
161, 164, 165, 187, 192, 208, 209
Time diary, 179
Social Exclusion Unit (SEU), 3, 6, 60, 61, 82,
Total Difficulties Score of the Strengths and
85, 97
Difficulties Instrument (SDQ), 72,
Social grandparenting, 175
73, 74
Social housing, 34, 100
Total fertility rate, 2, 48, 81
Social norms, 99, 106, 107 Traditional attitudes, 28, 109, 165
Social polarisation, 3, 61, 101 Transnational care, 13
Social policy, 177 Travel costs, 211
Social security, 10 Trial marriage, 115, 118, 120
Socio-economic status, 131, 132, 134, 135, Triplet, 69
136, 139, 141, 142, 144 Twin, 61, 69
Solo living, 25
Spain, 128, 129, 130, 132, 174, 178, 180, 185 U
Special educational needs, 15, 219 UK Household Longitudinal Study, 180
Special Migration Statistics (SMS), 14, 192, Ukraine, 132, 136
193, 194, 195, 197, 198, 201 UK Time Use Survey, 183
Special Travel Statistics (STS), 193 Unemployment, 25, 65, 66, 89, 90, 91,
Special Workplace Statistics (SWS), 193 92, 212
Spouse, 10, 44, 127, 128, 131, 132, 133, 134, Unhappiness, 9, 136, 145
135, 136, 137, 138, 139, 140, 141, 142, Union
143, 144, 155, 171, 183, 184 creation, 8
Statutory leave, 176 dissolution, 8
Stepdaughter, 155 formation, 25, 105, 106
Stepfamily, 151, 152, 153, 154, 155, 157, 159, instability, 111
160, 161, 162, 163, 164, 165, 166, 167 United Kingdom, 1, 11, 14, 26, 41, 81,
Stepfather, 152, 155 186, 191
Stepmother, 153–155, 158–160 University student, 14–15
Index 245
Unmarried, 8, 10, 26, 105, 108, 127, 129, 132, Wages, 27, 43, 44, 45, 76
134, 136, 138, 139, 141, 142, 143, 144, Wales, 10, 12, 15, 25, 26, 47, 62, 63, 64, 71,
145, 152 105, 120, 128, 130, 131, 132, 133, 134,
Unpaid care, 12 135, 138–143, 144, 145, 193, 194
Unpaid childcare, 171 Wedding, 120
Unpaid work, 179 Welfare
Unplanned infertility, 62 regime, 10, 128, 131, 144
Unplanned pregnancy, 62, 97, 100
state, 5, 11, 25, 172, 176, 177
USA, 111, 129, 130, 139, 177
Well-being, 9, 10, 56, 127–146, 151, 152, 154,
U.S. Health and Retirement Study (HRS), 180
155, 177
V Widow/hood, 28, 114, 132, 134, 141, 151
Vital registration, 131 Women, 3, 4, 5, 6, 7, 10, 12, 24, 25, 26, 28–37,
Voluntary-aided schools (VA), 217, 218, 219, 41–57, 59, 61, 62, 63, 64, 65, 66, 67,
231, 232, 233, 234 69, 75, 76, 77, 78, 81, 83, 85, 87, 97,
Voluntary-controlled schools (VC), 217, 218, 99, 100, 113, 115, 116, 117, 118, 119,
231, 232 120, 127, 129, 130, 134, 135, 136, 137,
Volunteer work, 172 138, 139, 140, 141, 142, 143, 144, 145,
Vulnerability, 129 152, 153, 160, 171, 172, 174, 177, 185,
186, 187
W Work/life policy, 77–78, 96–97, 107–112,
Wage dispersion, 45 133–144, 177–178

Fertility, Living Arrengements, Care and Mobilitie - Hohn Stillwell, Ernestina Coast, Dylan Kneale

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Fertility, Living Arrengements, Care and Mobilitie - Hohn Stillwell, Ernestina Coast, Dylan Kneale

Загружено:

Авторское право:

Доступные форматы

Fertility, Living Arrangements, Care and Mobility

Understanding Population Trends and Processes

For further volumes:

Understanding Population Trends

c Springer Science + Business Media B.V. 2009

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Swindon Ian Diamond

Leeds, UK John Stillwell

1 Fertility, Living Arrangements, Care and Mobility . . . . . . . . . . . . . . . . . 1

2 Delayed Childbearing and Childlessness . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3 Women’s Education and Childbearing: A Growing Divide . . . . . . . . . . 41

4 The Timing of Motherhood, Mothers’ Employment and Child

5 Early Parenthood: Definition and Prediction in Two British Cohorts . 81

6 Currently Cohabiting: Relationship Attitudes, Expectations and

7 Living Arrangements, Health and Well-Being . . . . . . . . . . . . . . . . . . . . . . 127

8 Stepparenting and Mental Health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

9 Grandparents and the Care of Their Grandchildren . . . . . . . . . . . . . . . . 171

10 Internal Migration and Inter-household Relationships . . . . . . . . . . . . . . 191

11 Exploring Dimensions of School Change During Primary Education

Paul Boyle School of Geography and Geosciences, University of St Andrews, St

Ernestina Coast Department of Social Policy, London School of Economics,

Oliver Duke-Williams School of Geography, University of Leeds, Leeds LS2

Peteke Feijten School of Geography and Geosciences, University of St Andrews,

Zhiqiang Feng School of Geography and Geosciences, University of St Andrews,

Vernon Gayle Department of Applied Social Science, University of Stirling,

Elspeth Graham School of Geography and Geosciences, University of St

Kirstine Hansen Centre for Longitudinal Studies, Institute of Education,

Denise Hawkes Business School, University of Greenwich, London, UK and

Heather Joshi Centre for Longitudinal Studies, Institute of Education, University

Dylan Kneale Centre for Longitudinal Studies, Institute of Education, University

Dylan Kneale, Ernestina Coast and John Stillwell

Fig. 1.1 Age-specific 60 29.5

Births per 1,000 women

Average age (Years)

Perspectives about whether change is beneficial or otherwise to society can

Table 1.1 Old age dependency ratio, 2005–2050, selected countries

Belgium 26.3 36.5 48.1 82.9

Source: Eurostat (2004) based on the assumption that net immigration

International migration has received considerable attention in the academic and

Fig. 1.2 Proportions of 90

cycle – impacting on different members of the group. Precise numbers of individu-

Table 1.2 Main data sets used in forthcoming chapters

Chapter Author(s) Data sources

2 Simpson National Child Development Study

than a manifestation of change in birth intervals (Sobotka, 2004). There is some

Fig. 2.1 Completed family 25

Data and Methods

The Propensity to Childlessness Among Young

Delayed Childbearing by Sex

Delayed Childbearing by Social Class Background

Fig. 2.3 Childlessness at 33 by social class background (%)

Table 2.1 Men and women remaining childless at 33 by cohort

Social class background (father’s occupation at age 16)

Source: Author’s calculations based on survey data.

Childlessness by Educational Attainment

Recent decades have seen a massive expansion in post-secondary education, with

Fig. 2.4 Childlessness at 33 by educational qualifications (%)

Childlessness by Current Economic Activity

Childlessness by Partnership Status

in terms of both subjectivities as well as practices, in order to understand fertility

Sarah Smith and Anita Ratcliffe

Proportion with qualifications

Fig. 3.1 Highest educational qualification, by date of birth cohort